Research ArticleDesign and Simulation of Human Resource Allocation ModelBased on Double-Cycle Neural Network
Qi Feng 1 Zixuan Feng2 and Xingren Su1
1Panzhihua University Panzhihua 617000 Sichuan China2Hongkong Shue Yan University Hongkong 999077 China
Correspondence should be addressed to Qi Feng fengqipzhueducn
Received 26 August 2021 Revised 14 September 2021 Accepted 17 September 2021 Published 25 October 2021
Academic Editor Bai Yuan Ding
Copyright copy 2021Qi Feng et alis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
e rationalization of human resource management is helpful for enterprises to efficiently train talents in the field improve themanagement mode and increase the overall resource utilization rate of enterprises e current computational models applied inthe field of human resources are usually based on statistical computation which can no longer meet the processing needs ofmassive data and do not take into account the hidden characteristics of data which can easily lead to the problem of informationscarcity e paper combines recurrent convolutional neural network and traditional human resource allocation algorithm anddesigns a double recurrent neural network job matching recommendation algorithm applicable to the human resource fieldwhich can improve the traditional algorithm data training quality problem In the experimental part of the algorithm thearithmetic F1 value in the paper is 0823 which is 201 and 74 higher than the other two algorithms respectively indicatingthat the algorithm can improve the hidden layer features of the data and then improve the training quality of the data and improvethe job matching and recommendation accuracy
1 Introduction
With the development of artificial intelligence technologythe intelligent process of enterprises is also advancingrapidly which is reflected in two aspects of managementintelligence and equipment automation [1] As an importantpart of management intelligence intelligent management ofhuman resources is being paid more and more attention bymore and more enterprise organizations [2 3] e rationalmanagement of human resources is of great help to en-terprises in efficiently cultivating talents in the field of en-terprises perfecting the management mode of enterprisesand improving the overall resource utilization rate of en-terprises which can effectively strengthen the integrity ofenterprises and at the same time give better play to theeconomic and social value of enterprises [4] erefore it isespecially important to carry out efficient and intelligenthuman resource management for enterprises
However the distinctive feature of the intelligent era isthe explosive growth of data which makes the traditional
human resource management methods no longer able tomeet the massive data processing needs e simple HRmanagement system cannot make accurate analysis anddeployment of the enterprisersquos manpower data and alsowastes a lot of data collected by various information systemsof the enterprise [5] is not only loses the enterprisersquosinformation resources but also slows down the intelligentinformation process erefore it is necessary to applyadvanced artificial intelligence algorithms to the enterprisehuman resource allocation system which can significantlyimprove the processing capability of enterprise human re-source data [6 7]
Fundamentally the processing of data is data miningData mining refers to the use of software systems to rea-sonably extract the useful information contained in the dataand most traditional data mining methods use statisticalmethods [8] Statistical methods can be used in the case ofsmall data content but if there is a large amount of datastatistical methods are less able to adapt to the need Ma-chine learning is a widely used data mining method that can
HindawiComputational Intelligence and NeuroscienceVolume 2021 Article ID 7149631 10 pageshttpsdoiorg10115520217149631
be trained on large amounts of data and extract the hiddenfeatures of the data [9] ese features are then continuouslylearned and data information extraction can be performedaccurately In this paper machine learning method is used toprocess human resource data and then improve the humanresource data processing capability
is paper focuses on the optimization of the compo-sition of the human resources structure in terms of per-sonnel categories which is to accurately forecast the futurehuman resources demand for each department and eachtype of position for the enterprise in advance so as toimprove the recruitment efficiency avoid human resourcesshortage and guide human resources training thus main-taining the stable operation of the enterprise and preventingbusiness risks [10] Forecasting is the estimation and ex-trapolation of the future which studies the future devel-opment of things and their operational rules and estimatesand analyzes the trend of changes of its various elements Inorder to achieve this purpose it is often necessary to imitateor abstract the real world (object) and this process is calledmodeling a representation and embodiment of the realworld (object) obtained by means of modeling is called amodel [11] All objectively existing things and theirmovement forms are collectively called reality reality andthe future are not the same but the future can be foreseenthrough the study of reality which is prediction From theperspective of information movement reality contains thefuture and nurtures the future erefore a ldquogoodrdquo modelshould not only reflect reality but also accurately predict thefuture development us it is necessary to build a math-ematical model that is in line with the objective developmentof things to make predictions
Human resource demand recommendation is a humanresource planning activity that takes the organizationrsquosestablished goals development plans and work tasks as thestarting point and takes into account the influence of variousinternal and external factors to forecast the quantity qualityand structure of human resources required by the organi-zation in a certain period in the future [12 13] Unliketraditional mathematical modeling methods neural net-works have the ability to simulate part of human imaginativethinking and find out the characteristic relationship(mapping) between the input (influencing factors) and theoutput (human resource requirements) through learningand memory association of historical data In the artificialneural network the explanatory variables in the historicalsample data can be used as the input units of the neuralnetwork and the output units are obtained after the op-eration of the neural network implicit layer weights andactivation functions [14] e objective function is selectedie the appropriate neural network weights are chosen tominimize the sum of squares of the difference between thedesired output and the actual output of the neural networkroughmultisample learning the weights are modified andthe deviation is continuously reduced so that the explanatoryvariables are optimally fitted to the explanatory variablesand the new known explanatory variables are input into the
neural network and the predicted values are output throughthe implicit layer [15]
Since the human resource structure of an enterprise isa function of social economic political and techno-logical factors modeling by conventional mathematicalmethods is not only a large workload but also difficult toguarantee accuracy [16] Recurrent neural network hasstrong ability of nonlinear learning and pattern recog-nition through which the relationship between humanresource structure and its influencing factors can bemodeled with relatively small error and high accuracy[17] Neural networks have been used in a large numberof applications in fiscal forecasting management decisionmaking and process control On the one hand we candraw inspiration from the experience of neural networkapplications in other fields on the other hand we arerequired to design and develop new two-loop neuralnetwork models and algorithms in order to solve theproblems of market research
2 Human Resource Allocation Model IsCombined with Neural Network
21 Traditional Human Resource Allocation ModelAccording to the traditional human resource allocationtheory planning human resources mainly involves analyz-ing the personnel structure of the unit and sorting out thecorrelation between job requirements and personnel com-petencies in detail Personnel competencies include variouselements which are weighted and summed to determine thequality score of personnel [18 19]
e traditional HR scoring process is shown in Figure 1First the incoming data are grouped and analyzed into twogroups the personnel evaluation matrix and the personnelcompetency matrix [20] e most commonly used is theemployee competency matrix which may take into accountvarious factors such as self-evaluation superior-subordinateevaluation and patient evaluation e personnel compe-tency matrix includes information such as employee per-formance attendance and job title After obtaining thepersonnel evaluation matrix values and the personnelcompetency matrix values the key indicator job match canbe obtained as followswhere n1 sim n4 is the correspondingevaluation parameter
erefore the personnel can be optimized by the jobmatching model as shown in the following equation
2 Computational Intelligence and Neuroscience
maxZ 1113944n
i11113944
m
j1sijXij
1113944
n
i1Xij 1 j 1 2 m
1113944
m
j1Xij le 1 i 1 2 n
(3)
e above algorithm is simple and effective and allowsfor a better allocation of human resources but this is onlyapplicable when there is little manpower data As the
enterprise system increases the manpower data also in-crease and the problem becomes complicated is methodis inefficient in calculation and does not allow for better datamining and effective management of human resources
22 Improved Recurrent Neural Network Model From theabove we can see that the essence of HR scheduling model isto analyze HR data and calculate the job matching scoreen the scheduling of personnel is based on the jobmatching score which can be abstracted as a recommen-dation model in essence Recommendation models havebeen analyzed and validated in many fields and the currentmainstream recommendation models use recurrent neuralnetworks as the data processing module [21]
emost important feature of recurrent neural networksis the use of recurrent convolution for data training oper-ations [22 23] e convolutional recurrent network modelcan be regarded as a hierarchical data model and the inputof the convolutional network is the original human resourcedata Abstract features between the data are extractedthrough the process of recurrent convolution operationpooling and activation function and the process isexpressed as follows
x1⟶ ω1⟶ x
2⟶ middot middot middot⟶ xLminus1⟶ ωLminus1⟶ x
L⟶ ωL⟶ z
z f xL y1113872 1113873
(4)
In this paper xL is the data input of L layer ω is theparameter weight value of L layer z is the loss functionselected by the model y is the calibration value of the modeland the function f is the final calculation parameter of themodel In this paper the basic neural network is improvedby using a hybrid recurrent neural network model and theglobal model is combined with the local model to process thedata by using the data features of the hierarchical model asthe network output e hierarchical model structure is thenused to build the network and realize the job matchingrecommendation [24 25] e hybrid recurrent networkmodel is shown in Figure 2
In the process of model building cross entropy [26] isselected as the loss functions in the paper is loss functioncan compare the actual value of the data with the expectedvalue of the data and then determine the closeness of thedata and the loss function is as follows
L minus 1113944n
(y log p +(1 minus y)log(1 minus p)) (5)
At the same time the parameters are optimized duringtraining using a gradient optimization algorithm is waythe parameter transfer can be as accurate as possible and thespecific update process of the model parameters is thefollowing
(1) e learning rate and the number of iterations of theneural network are updated as shown in the fol-lowing equation
mt β1mtminus1 + 1 minus β1( 1113857gt
vt β2vtminus1 + 1 minus β2( 1113857gt(6)
where β1 β2 are the hypermastigote of the recurrentneural network gt is the computational gradient ofthe model and t is the number of iterations of themodel
(2) Optimal orientation of the first-order and second-order estimates [27] is
1113954mt mt
1 minus βt1
1113954vt vt
1 minus βt2
(7)
(3) Update the parameters of the model from the resultsobtained above
θt+1 θt minusl 1113954mt1113954vt
1113968+ ε
(8)
Data processing
Data grouping
Statistical analysis
Optimization calculation
Person post matching value
Assign output
Figure 1 HR scoring process
Computational Intelligence and Neuroscience 3
3 Algorithm Flow
As mentioned in Section 22 the overall algorithm used isthe job matching recommendation algorithm e processdesign of the recommendation algorithm should also betailored to different application scenarios and data char-acteristics Currently used algorithms in the HR field areusually statistical algorithms that do not take into accountthe hidden features of the data and rely on simple scoringand expert judgment mechanisms which can easily lead toinformation scarcity problems [28]
In this paper we combine recurrent CNN and tradi-tional human resource allocation algorithm to design a jobmatching recommendation algorithm for human resourcefield e algorithm not only improves the problem of lowdata training quality of traditional algorithms but also ef-fectively improves the data computation efficiency by usingrecurrent neural networks [29] e core idea of the algo-rithm is to first extract the original features from the datawhich are consistent with the features required by traditionalhuman resources including the personnel evaluation matrixand the personnelrsquos ability matrixe data are extracted andencoded in an encoder and the encoded features are then fedas data input to the data input layer of RNN e data areprocessed using the recurrent convolutional layer to obtainthe job matching results and the algorithm flow is shown inFigure 3
First the data is collected and selected using a distributedstreaming collection method and grouped into a personnelevaluation matrix and a personnel competency matrix edata are then abstracted preprocessed encoded using anencoder and saved to a data warehouse and further enhancedusing a feature enhancement algorithm and fed into a RNN
e final job matching score is output and the HR rec-ommendation process is completed
e steps in the algorithm flow are described as follows
(1) Data collection using the distributed streaming datacollection method the format of manpower datavaries from company to company ereforemanpower data must be processed in a uniformformat including data rounding and conversionoperations
(2) Preprocessing of the raw data the datum is groupedso that a more comprehensive understanding of theHR model characteristics can be obtained e da-tum is also saved to the data warehouse so thatsubsequent data model training can be supported
(3) Perform feature enhancement the datum is obtainedfrom the data warehouse and the results of datagrouping are learned Fused data results are used asneural network data input for network training
(4) Recommendation result output the job matchingresults are ranked and then the reasonable job as-signment is made with reference to the score
31 Algorithm Evaluation Metrics e algorithms areevaluated by a certain number of metrics and the accurate
Input feature value
Channel 1 Channel n
Data splicing layer
Convolution Local connection
Averaging
Recommended results
Figure 2 Double-loop network model
Data acquisition
Capability evaluation
Data preprocessing
Data warehouse
Extract feature information
Enhanced feature model
Cyclic neural network
Personnel evaluation
User information
Person post matching score
Output results
Statistical information
Construct scoring matrix
Start
Figure 3 Algorithm flow
4 Computational Intelligence and Neuroscience
evaluation metrics are used in the evaluation of humanresource recommendation algorithms in this paper [13 22]
In the evaluation system the higher the accuracy andrecall the better the algorithm However in some scenariosaccuracy and recall may be contradictory so in order tosynthesize these two metrics the F1 value [30] is used in thepaper for synthesis and the formula is
precision tr
tr + fr
recall tr
tr + fnr
F1 2 middot precision middot recallprecision + recall
(9)
4 Experiment and Analysis
41 Experimental Data
411 Data 1 e data in this paper are sorted into threetypes of enterprise human resource data including per-sonnel information personnel evaluation matrix values andpersonnel capability matrix values e datum was collectedfrom 4560 employees and 1233 positions with a sample sizeof 134540 [31]
412 Data 2 We included a large power supply company ina certain region the details of the company are as followscompany A is the main power supply and management unitin the region and the power supply covers 29 towns and 2forest farms under the jurisdiction of the region with a totalof 31 secondary power supply companies the company hasbeen established for more than 30 years and after a longperiod of development the companyrsquos staff structure andbusiness scope (service area) and quantity (mainly electricitysales and equipment) have changed dramatically By 2020more than 97 of the companyrsquos existing substations will beunmanned and the service will be based on a network ofeight 500 kV substations With the deepening of market-oriented reform of the national power grid the developmentof the company is facing new challenges with managementmethods needing improvement personnel structure need-ing optimization and power supply equipment and facilitiesneeding further upgrading In this context Company A hasput forward the human resources slogan of ldquopositive changetalent firstrdquo and strives to build a human resources man-agement system to adapt to the new situation striving tobecome a first-class power supply company in China and aleading player in the industry Information on the com-panyrsquos human resources and company performance ismainly summarized through the companyrsquos informationrelease [32]
Figure 4 depicts the trends of the number of employeesand electricity sales of company A from 2010 to 2020 It canbe seen that the number of employees of company A hasbeen maintaining a growth trend with a relatively slowgrowth rate and the electricity sales of company A except
for a certain downward trend around 2015 have generallymaintained a good growth and still maintained a goodgrowth in the recent year of total growth for the decline inelectricity demand during 2015 may be caused by factorssuch as the weakness of the domestic economy
Although the number of employees of the company hasmaintained growth the adjustment of relevant nationalpolicies and changes in the companyrsquos internal human re-sources structure have led to certain problems in the com-panyrsquos personnel structure which is manifested in the highdemand for power dispatching and transmission and sub-station personnel in the main power transmission network Inthis regard it is necessary for the company to plan its humanresources department in advance complete the human re-sources demand forecast as early as possible adjust andoptimize the personnel structure clean up the surplus per-sonnel and introduce the insufficient number of professionalsto serve the companyrsquos long-term development goals
rough the above basic situation it can be seen that theinformation related to the development history of companyA is relatively detailed and rich in data and the collected datashows certain volatility which is suitable for the analysis andprediction of its human resource demand by using double-loop neural network
42 Data 1 Experimental Test and Analysis of Results In thispaper the samples are divided into a training sample set anda test sample set e pseudocode of the neural networktesting procedure is given in Algorithm 1
en the feasibility of the algorithm in the paper iscompared with the experiments using the algorithm in thepaper CNN and the traditional statistical method for modeltraining and experimental analysis and the analysis evalu-ation indexes are accuracy recall and F1 value e ex-perimental results are shown in Table 1
times104times104
Electricity salesTotal personnel
0
1
2
3
4
5
6
2
22
24
26
28
3
32
34
36
38
4
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2010
Figure 4 Trend of the number of employees and electricity sales ofcompany A from 2010 to 2020
Computational Intelligence and Neuroscience 5
As can be seen from Table 1 the F1 value of the algo-rithm in the paper is significantly improved compared withthe other two algorithms and the traditional statisticalmethod is the least effective F1 value is 0678 in the case oflarge amount of data and the F1 value of common CNN isimproved to 0766 the best performance of the algorithm inthe paper is 0823 is indicates that the direct use of re-current convolutional network does not really improve thetraining features of the data while the method of usingglobal network plus local network in the paper can effectivelyimprove the hidden layer features of the data and thenimprove the data training quality and improve the matchingdegree and recommendation accuracy of the algorithm
43 Model Analytical Indexes and Screening in Data 2Experiments In this paper we established the analysis indexsystem of enterprise A and screened all the indexes based onthe gray correlation method to determine the final keyanalytical indexes of human resources demand
Based on the above principles the following analysisindexes were selected as the initial analysis system of HRdemand forecast for company A e specific sinks areshown in Table 2
After the construction of the original demand fore-casting analysis system we need to screen the indicators thataffect the human resource demand of company A ie thekey indicators e screening method is the gray correlationanalysis method of calculating the comprehensive correla-tion degree and the analysis results are summarized inTable 3 e correlation between each indicator and humanresources is calculated by the gray systemmodeling softwareand the specific results are as follows
ρa1 07896
ρa2 06157
ρa3 06875
ρa4 05792
ρa5 05680
ρa6 06080
(10)
e higher the value of the correlation the greater theinfluence on human resources demand By ranking the abovecorrelations we can get a1 a2 a3 a4 four variables withrelatively large values ie the length of transmission network
Input feature DOutput hybrid recurrent neural modelInitialize the hypermastigote which include the number of iterations t the learning rate L the hypermastigote of the recurrent neuralnetwork β1 β2 and the computational gradient of the model gt(1) i cycles from 1 to t(2) j cycles from 1 to t(3) Calculate the eigenvalues of each channel and substitute them into the function f(4) If j t then terminate the loop and execute the step 1(5) If jlt t go back to step 1(6) Extracting convolutional features to obtain F(7) Combine the F1 values and local model eigenvalues to obtain the probability values(8) Get the current job match value(9) Sort and output the final result(10) If ilt t then return to step (2) and loop through the i process(11) If i t end
ALGORITHM1 e neural network testing procedure
Table 1 Comparative experimental result
Model Accuracy Recall F1Model in this paper 08122 0832 0823Statistical model 0702 0649 0678Cyclic neural network 0755 0775 0766
Table 2 Company Arsquos human resource demand forecasts analysis system
Category 1 analysis index Category 2 analysis indicators Specific analysis indicators
Company development resources Core resources owned by the company Transmission network length aNumber of substations under jurisdiction a
Market development of the company Company market size Total number of users aTotal population served
Company development objectives Company size Annual revenue of the companyOperating conditions of the company Online electricity sales
6 Computational Intelligence and Neuroscience
the number of substations the total number of customers andthe amount of feed-in tariffs have a greater influence on theanalysis of human resources demand so the above fouranalysis indicators are selected as key indicators [33]
44Data 2Forecast e implementation of the double-loopneural network model consists of two stages first the keyindex prediction values are input into model to obtain thefinal prediction values
e predicted values of key analytical indicators ofhuman resource requirements of company A 2017ndash2019 areshown in Table 4
Figure 5 shows the error trend of the neural networktraining It can be seen that the output values obtained fromthe network training do not differ much from the optimaloutput valuese results are shown in Figure 6 R 09999 isobtained which initially indicates that our model is welltrained
Table 3 Summary of the results of the analysis of key indicators of human resources needs of enterprise A
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
be trained on large amounts of data and extract the hiddenfeatures of the data [9] ese features are then continuouslylearned and data information extraction can be performedaccurately In this paper machine learning method is used toprocess human resource data and then improve the humanresource data processing capability
is paper focuses on the optimization of the compo-sition of the human resources structure in terms of per-sonnel categories which is to accurately forecast the futurehuman resources demand for each department and eachtype of position for the enterprise in advance so as toimprove the recruitment efficiency avoid human resourcesshortage and guide human resources training thus main-taining the stable operation of the enterprise and preventingbusiness risks [10] Forecasting is the estimation and ex-trapolation of the future which studies the future devel-opment of things and their operational rules and estimatesand analyzes the trend of changes of its various elements Inorder to achieve this purpose it is often necessary to imitateor abstract the real world (object) and this process is calledmodeling a representation and embodiment of the realworld (object) obtained by means of modeling is called amodel [11] All objectively existing things and theirmovement forms are collectively called reality reality andthe future are not the same but the future can be foreseenthrough the study of reality which is prediction From theperspective of information movement reality contains thefuture and nurtures the future erefore a ldquogoodrdquo modelshould not only reflect reality but also accurately predict thefuture development us it is necessary to build a math-ematical model that is in line with the objective developmentof things to make predictions
Human resource demand recommendation is a humanresource planning activity that takes the organizationrsquosestablished goals development plans and work tasks as thestarting point and takes into account the influence of variousinternal and external factors to forecast the quantity qualityand structure of human resources required by the organi-zation in a certain period in the future [12 13] Unliketraditional mathematical modeling methods neural net-works have the ability to simulate part of human imaginativethinking and find out the characteristic relationship(mapping) between the input (influencing factors) and theoutput (human resource requirements) through learningand memory association of historical data In the artificialneural network the explanatory variables in the historicalsample data can be used as the input units of the neuralnetwork and the output units are obtained after the op-eration of the neural network implicit layer weights andactivation functions [14] e objective function is selectedie the appropriate neural network weights are chosen tominimize the sum of squares of the difference between thedesired output and the actual output of the neural networkroughmultisample learning the weights are modified andthe deviation is continuously reduced so that the explanatoryvariables are optimally fitted to the explanatory variablesand the new known explanatory variables are input into the
neural network and the predicted values are output throughthe implicit layer [15]
Since the human resource structure of an enterprise isa function of social economic political and techno-logical factors modeling by conventional mathematicalmethods is not only a large workload but also difficult toguarantee accuracy [16] Recurrent neural network hasstrong ability of nonlinear learning and pattern recog-nition through which the relationship between humanresource structure and its influencing factors can bemodeled with relatively small error and high accuracy[17] Neural networks have been used in a large numberof applications in fiscal forecasting management decisionmaking and process control On the one hand we candraw inspiration from the experience of neural networkapplications in other fields on the other hand we arerequired to design and develop new two-loop neuralnetwork models and algorithms in order to solve theproblems of market research
2 Human Resource Allocation Model IsCombined with Neural Network
21 Traditional Human Resource Allocation ModelAccording to the traditional human resource allocationtheory planning human resources mainly involves analyz-ing the personnel structure of the unit and sorting out thecorrelation between job requirements and personnel com-petencies in detail Personnel competencies include variouselements which are weighted and summed to determine thequality score of personnel [18 19]
e traditional HR scoring process is shown in Figure 1First the incoming data are grouped and analyzed into twogroups the personnel evaluation matrix and the personnelcompetency matrix [20] e most commonly used is theemployee competency matrix which may take into accountvarious factors such as self-evaluation superior-subordinateevaluation and patient evaluation e personnel compe-tency matrix includes information such as employee per-formance attendance and job title After obtaining thepersonnel evaluation matrix values and the personnelcompetency matrix values the key indicator job match canbe obtained as followswhere n1 sim n4 is the correspondingevaluation parameter
erefore the personnel can be optimized by the jobmatching model as shown in the following equation
2 Computational Intelligence and Neuroscience
maxZ 1113944n
i11113944
m
j1sijXij
1113944
n
i1Xij 1 j 1 2 m
1113944
m
j1Xij le 1 i 1 2 n
(3)
e above algorithm is simple and effective and allowsfor a better allocation of human resources but this is onlyapplicable when there is little manpower data As the
enterprise system increases the manpower data also in-crease and the problem becomes complicated is methodis inefficient in calculation and does not allow for better datamining and effective management of human resources
22 Improved Recurrent Neural Network Model From theabove we can see that the essence of HR scheduling model isto analyze HR data and calculate the job matching scoreen the scheduling of personnel is based on the jobmatching score which can be abstracted as a recommen-dation model in essence Recommendation models havebeen analyzed and validated in many fields and the currentmainstream recommendation models use recurrent neuralnetworks as the data processing module [21]
emost important feature of recurrent neural networksis the use of recurrent convolution for data training oper-ations [22 23] e convolutional recurrent network modelcan be regarded as a hierarchical data model and the inputof the convolutional network is the original human resourcedata Abstract features between the data are extractedthrough the process of recurrent convolution operationpooling and activation function and the process isexpressed as follows
x1⟶ ω1⟶ x
2⟶ middot middot middot⟶ xLminus1⟶ ωLminus1⟶ x
L⟶ ωL⟶ z
z f xL y1113872 1113873
(4)
In this paper xL is the data input of L layer ω is theparameter weight value of L layer z is the loss functionselected by the model y is the calibration value of the modeland the function f is the final calculation parameter of themodel In this paper the basic neural network is improvedby using a hybrid recurrent neural network model and theglobal model is combined with the local model to process thedata by using the data features of the hierarchical model asthe network output e hierarchical model structure is thenused to build the network and realize the job matchingrecommendation [24 25] e hybrid recurrent networkmodel is shown in Figure 2
In the process of model building cross entropy [26] isselected as the loss functions in the paper is loss functioncan compare the actual value of the data with the expectedvalue of the data and then determine the closeness of thedata and the loss function is as follows
L minus 1113944n
(y log p +(1 minus y)log(1 minus p)) (5)
At the same time the parameters are optimized duringtraining using a gradient optimization algorithm is waythe parameter transfer can be as accurate as possible and thespecific update process of the model parameters is thefollowing
(1) e learning rate and the number of iterations of theneural network are updated as shown in the fol-lowing equation
mt β1mtminus1 + 1 minus β1( 1113857gt
vt β2vtminus1 + 1 minus β2( 1113857gt(6)
where β1 β2 are the hypermastigote of the recurrentneural network gt is the computational gradient ofthe model and t is the number of iterations of themodel
(2) Optimal orientation of the first-order and second-order estimates [27] is
1113954mt mt
1 minus βt1
1113954vt vt
1 minus βt2
(7)
(3) Update the parameters of the model from the resultsobtained above
θt+1 θt minusl 1113954mt1113954vt
1113968+ ε
(8)
Data processing
Data grouping
Statistical analysis
Optimization calculation
Person post matching value
Assign output
Figure 1 HR scoring process
Computational Intelligence and Neuroscience 3
3 Algorithm Flow
As mentioned in Section 22 the overall algorithm used isthe job matching recommendation algorithm e processdesign of the recommendation algorithm should also betailored to different application scenarios and data char-acteristics Currently used algorithms in the HR field areusually statistical algorithms that do not take into accountthe hidden features of the data and rely on simple scoringand expert judgment mechanisms which can easily lead toinformation scarcity problems [28]
In this paper we combine recurrent CNN and tradi-tional human resource allocation algorithm to design a jobmatching recommendation algorithm for human resourcefield e algorithm not only improves the problem of lowdata training quality of traditional algorithms but also ef-fectively improves the data computation efficiency by usingrecurrent neural networks [29] e core idea of the algo-rithm is to first extract the original features from the datawhich are consistent with the features required by traditionalhuman resources including the personnel evaluation matrixand the personnelrsquos ability matrixe data are extracted andencoded in an encoder and the encoded features are then fedas data input to the data input layer of RNN e data areprocessed using the recurrent convolutional layer to obtainthe job matching results and the algorithm flow is shown inFigure 3
First the data is collected and selected using a distributedstreaming collection method and grouped into a personnelevaluation matrix and a personnel competency matrix edata are then abstracted preprocessed encoded using anencoder and saved to a data warehouse and further enhancedusing a feature enhancement algorithm and fed into a RNN
e final job matching score is output and the HR rec-ommendation process is completed
e steps in the algorithm flow are described as follows
(1) Data collection using the distributed streaming datacollection method the format of manpower datavaries from company to company ereforemanpower data must be processed in a uniformformat including data rounding and conversionoperations
(2) Preprocessing of the raw data the datum is groupedso that a more comprehensive understanding of theHR model characteristics can be obtained e da-tum is also saved to the data warehouse so thatsubsequent data model training can be supported
(3) Perform feature enhancement the datum is obtainedfrom the data warehouse and the results of datagrouping are learned Fused data results are used asneural network data input for network training
(4) Recommendation result output the job matchingresults are ranked and then the reasonable job as-signment is made with reference to the score
31 Algorithm Evaluation Metrics e algorithms areevaluated by a certain number of metrics and the accurate
Input feature value
Channel 1 Channel n
Data splicing layer
Convolution Local connection
Averaging
Recommended results
Figure 2 Double-loop network model
Data acquisition
Capability evaluation
Data preprocessing
Data warehouse
Extract feature information
Enhanced feature model
Cyclic neural network
Personnel evaluation
User information
Person post matching score
Output results
Statistical information
Construct scoring matrix
Start
Figure 3 Algorithm flow
4 Computational Intelligence and Neuroscience
evaluation metrics are used in the evaluation of humanresource recommendation algorithms in this paper [13 22]
In the evaluation system the higher the accuracy andrecall the better the algorithm However in some scenariosaccuracy and recall may be contradictory so in order tosynthesize these two metrics the F1 value [30] is used in thepaper for synthesis and the formula is
precision tr
tr + fr
recall tr
tr + fnr
F1 2 middot precision middot recallprecision + recall
(9)
4 Experiment and Analysis
41 Experimental Data
411 Data 1 e data in this paper are sorted into threetypes of enterprise human resource data including per-sonnel information personnel evaluation matrix values andpersonnel capability matrix values e datum was collectedfrom 4560 employees and 1233 positions with a sample sizeof 134540 [31]
412 Data 2 We included a large power supply company ina certain region the details of the company are as followscompany A is the main power supply and management unitin the region and the power supply covers 29 towns and 2forest farms under the jurisdiction of the region with a totalof 31 secondary power supply companies the company hasbeen established for more than 30 years and after a longperiod of development the companyrsquos staff structure andbusiness scope (service area) and quantity (mainly electricitysales and equipment) have changed dramatically By 2020more than 97 of the companyrsquos existing substations will beunmanned and the service will be based on a network ofeight 500 kV substations With the deepening of market-oriented reform of the national power grid the developmentof the company is facing new challenges with managementmethods needing improvement personnel structure need-ing optimization and power supply equipment and facilitiesneeding further upgrading In this context Company A hasput forward the human resources slogan of ldquopositive changetalent firstrdquo and strives to build a human resources man-agement system to adapt to the new situation striving tobecome a first-class power supply company in China and aleading player in the industry Information on the com-panyrsquos human resources and company performance ismainly summarized through the companyrsquos informationrelease [32]
Figure 4 depicts the trends of the number of employeesand electricity sales of company A from 2010 to 2020 It canbe seen that the number of employees of company A hasbeen maintaining a growth trend with a relatively slowgrowth rate and the electricity sales of company A except
for a certain downward trend around 2015 have generallymaintained a good growth and still maintained a goodgrowth in the recent year of total growth for the decline inelectricity demand during 2015 may be caused by factorssuch as the weakness of the domestic economy
Although the number of employees of the company hasmaintained growth the adjustment of relevant nationalpolicies and changes in the companyrsquos internal human re-sources structure have led to certain problems in the com-panyrsquos personnel structure which is manifested in the highdemand for power dispatching and transmission and sub-station personnel in the main power transmission network Inthis regard it is necessary for the company to plan its humanresources department in advance complete the human re-sources demand forecast as early as possible adjust andoptimize the personnel structure clean up the surplus per-sonnel and introduce the insufficient number of professionalsto serve the companyrsquos long-term development goals
rough the above basic situation it can be seen that theinformation related to the development history of companyA is relatively detailed and rich in data and the collected datashows certain volatility which is suitable for the analysis andprediction of its human resource demand by using double-loop neural network
42 Data 1 Experimental Test and Analysis of Results In thispaper the samples are divided into a training sample set anda test sample set e pseudocode of the neural networktesting procedure is given in Algorithm 1
en the feasibility of the algorithm in the paper iscompared with the experiments using the algorithm in thepaper CNN and the traditional statistical method for modeltraining and experimental analysis and the analysis evalu-ation indexes are accuracy recall and F1 value e ex-perimental results are shown in Table 1
times104times104
Electricity salesTotal personnel
0
1
2
3
4
5
6
2
22
24
26
28
3
32
34
36
38
4
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2010
Figure 4 Trend of the number of employees and electricity sales ofcompany A from 2010 to 2020
Computational Intelligence and Neuroscience 5
As can be seen from Table 1 the F1 value of the algo-rithm in the paper is significantly improved compared withthe other two algorithms and the traditional statisticalmethod is the least effective F1 value is 0678 in the case oflarge amount of data and the F1 value of common CNN isimproved to 0766 the best performance of the algorithm inthe paper is 0823 is indicates that the direct use of re-current convolutional network does not really improve thetraining features of the data while the method of usingglobal network plus local network in the paper can effectivelyimprove the hidden layer features of the data and thenimprove the data training quality and improve the matchingdegree and recommendation accuracy of the algorithm
43 Model Analytical Indexes and Screening in Data 2Experiments In this paper we established the analysis indexsystem of enterprise A and screened all the indexes based onthe gray correlation method to determine the final keyanalytical indexes of human resources demand
Based on the above principles the following analysisindexes were selected as the initial analysis system of HRdemand forecast for company A e specific sinks areshown in Table 2
After the construction of the original demand fore-casting analysis system we need to screen the indicators thataffect the human resource demand of company A ie thekey indicators e screening method is the gray correlationanalysis method of calculating the comprehensive correla-tion degree and the analysis results are summarized inTable 3 e correlation between each indicator and humanresources is calculated by the gray systemmodeling softwareand the specific results are as follows
ρa1 07896
ρa2 06157
ρa3 06875
ρa4 05792
ρa5 05680
ρa6 06080
(10)
e higher the value of the correlation the greater theinfluence on human resources demand By ranking the abovecorrelations we can get a1 a2 a3 a4 four variables withrelatively large values ie the length of transmission network
Input feature DOutput hybrid recurrent neural modelInitialize the hypermastigote which include the number of iterations t the learning rate L the hypermastigote of the recurrent neuralnetwork β1 β2 and the computational gradient of the model gt(1) i cycles from 1 to t(2) j cycles from 1 to t(3) Calculate the eigenvalues of each channel and substitute them into the function f(4) If j t then terminate the loop and execute the step 1(5) If jlt t go back to step 1(6) Extracting convolutional features to obtain F(7) Combine the F1 values and local model eigenvalues to obtain the probability values(8) Get the current job match value(9) Sort and output the final result(10) If ilt t then return to step (2) and loop through the i process(11) If i t end
ALGORITHM1 e neural network testing procedure
Table 1 Comparative experimental result
Model Accuracy Recall F1Model in this paper 08122 0832 0823Statistical model 0702 0649 0678Cyclic neural network 0755 0775 0766
Table 2 Company Arsquos human resource demand forecasts analysis system
Category 1 analysis index Category 2 analysis indicators Specific analysis indicators
Company development resources Core resources owned by the company Transmission network length aNumber of substations under jurisdiction a
Market development of the company Company market size Total number of users aTotal population served
Company development objectives Company size Annual revenue of the companyOperating conditions of the company Online electricity sales
6 Computational Intelligence and Neuroscience
the number of substations the total number of customers andthe amount of feed-in tariffs have a greater influence on theanalysis of human resources demand so the above fouranalysis indicators are selected as key indicators [33]
44Data 2Forecast e implementation of the double-loopneural network model consists of two stages first the keyindex prediction values are input into model to obtain thefinal prediction values
e predicted values of key analytical indicators ofhuman resource requirements of company A 2017ndash2019 areshown in Table 4
Figure 5 shows the error trend of the neural networktraining It can be seen that the output values obtained fromthe network training do not differ much from the optimaloutput valuese results are shown in Figure 6 R 09999 isobtained which initially indicates that our model is welltrained
Table 3 Summary of the results of the analysis of key indicators of human resources needs of enterprise A
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
maxZ 1113944n
i11113944
m
j1sijXij
1113944
n
i1Xij 1 j 1 2 m
1113944
m
j1Xij le 1 i 1 2 n
(3)
e above algorithm is simple and effective and allowsfor a better allocation of human resources but this is onlyapplicable when there is little manpower data As the
enterprise system increases the manpower data also in-crease and the problem becomes complicated is methodis inefficient in calculation and does not allow for better datamining and effective management of human resources
22 Improved Recurrent Neural Network Model From theabove we can see that the essence of HR scheduling model isto analyze HR data and calculate the job matching scoreen the scheduling of personnel is based on the jobmatching score which can be abstracted as a recommen-dation model in essence Recommendation models havebeen analyzed and validated in many fields and the currentmainstream recommendation models use recurrent neuralnetworks as the data processing module [21]
emost important feature of recurrent neural networksis the use of recurrent convolution for data training oper-ations [22 23] e convolutional recurrent network modelcan be regarded as a hierarchical data model and the inputof the convolutional network is the original human resourcedata Abstract features between the data are extractedthrough the process of recurrent convolution operationpooling and activation function and the process isexpressed as follows
x1⟶ ω1⟶ x
2⟶ middot middot middot⟶ xLminus1⟶ ωLminus1⟶ x
L⟶ ωL⟶ z
z f xL y1113872 1113873
(4)
In this paper xL is the data input of L layer ω is theparameter weight value of L layer z is the loss functionselected by the model y is the calibration value of the modeland the function f is the final calculation parameter of themodel In this paper the basic neural network is improvedby using a hybrid recurrent neural network model and theglobal model is combined with the local model to process thedata by using the data features of the hierarchical model asthe network output e hierarchical model structure is thenused to build the network and realize the job matchingrecommendation [24 25] e hybrid recurrent networkmodel is shown in Figure 2
In the process of model building cross entropy [26] isselected as the loss functions in the paper is loss functioncan compare the actual value of the data with the expectedvalue of the data and then determine the closeness of thedata and the loss function is as follows
L minus 1113944n
(y log p +(1 minus y)log(1 minus p)) (5)
At the same time the parameters are optimized duringtraining using a gradient optimization algorithm is waythe parameter transfer can be as accurate as possible and thespecific update process of the model parameters is thefollowing
(1) e learning rate and the number of iterations of theneural network are updated as shown in the fol-lowing equation
mt β1mtminus1 + 1 minus β1( 1113857gt
vt β2vtminus1 + 1 minus β2( 1113857gt(6)
where β1 β2 are the hypermastigote of the recurrentneural network gt is the computational gradient ofthe model and t is the number of iterations of themodel
(2) Optimal orientation of the first-order and second-order estimates [27] is
1113954mt mt
1 minus βt1
1113954vt vt
1 minus βt2
(7)
(3) Update the parameters of the model from the resultsobtained above
θt+1 θt minusl 1113954mt1113954vt
1113968+ ε
(8)
Data processing
Data grouping
Statistical analysis
Optimization calculation
Person post matching value
Assign output
Figure 1 HR scoring process
Computational Intelligence and Neuroscience 3
3 Algorithm Flow
As mentioned in Section 22 the overall algorithm used isthe job matching recommendation algorithm e processdesign of the recommendation algorithm should also betailored to different application scenarios and data char-acteristics Currently used algorithms in the HR field areusually statistical algorithms that do not take into accountthe hidden features of the data and rely on simple scoringand expert judgment mechanisms which can easily lead toinformation scarcity problems [28]
In this paper we combine recurrent CNN and tradi-tional human resource allocation algorithm to design a jobmatching recommendation algorithm for human resourcefield e algorithm not only improves the problem of lowdata training quality of traditional algorithms but also ef-fectively improves the data computation efficiency by usingrecurrent neural networks [29] e core idea of the algo-rithm is to first extract the original features from the datawhich are consistent with the features required by traditionalhuman resources including the personnel evaluation matrixand the personnelrsquos ability matrixe data are extracted andencoded in an encoder and the encoded features are then fedas data input to the data input layer of RNN e data areprocessed using the recurrent convolutional layer to obtainthe job matching results and the algorithm flow is shown inFigure 3
First the data is collected and selected using a distributedstreaming collection method and grouped into a personnelevaluation matrix and a personnel competency matrix edata are then abstracted preprocessed encoded using anencoder and saved to a data warehouse and further enhancedusing a feature enhancement algorithm and fed into a RNN
e final job matching score is output and the HR rec-ommendation process is completed
e steps in the algorithm flow are described as follows
(1) Data collection using the distributed streaming datacollection method the format of manpower datavaries from company to company ereforemanpower data must be processed in a uniformformat including data rounding and conversionoperations
(2) Preprocessing of the raw data the datum is groupedso that a more comprehensive understanding of theHR model characteristics can be obtained e da-tum is also saved to the data warehouse so thatsubsequent data model training can be supported
(3) Perform feature enhancement the datum is obtainedfrom the data warehouse and the results of datagrouping are learned Fused data results are used asneural network data input for network training
(4) Recommendation result output the job matchingresults are ranked and then the reasonable job as-signment is made with reference to the score
31 Algorithm Evaluation Metrics e algorithms areevaluated by a certain number of metrics and the accurate
Input feature value
Channel 1 Channel n
Data splicing layer
Convolution Local connection
Averaging
Recommended results
Figure 2 Double-loop network model
Data acquisition
Capability evaluation
Data preprocessing
Data warehouse
Extract feature information
Enhanced feature model
Cyclic neural network
Personnel evaluation
User information
Person post matching score
Output results
Statistical information
Construct scoring matrix
Start
Figure 3 Algorithm flow
4 Computational Intelligence and Neuroscience
evaluation metrics are used in the evaluation of humanresource recommendation algorithms in this paper [13 22]
In the evaluation system the higher the accuracy andrecall the better the algorithm However in some scenariosaccuracy and recall may be contradictory so in order tosynthesize these two metrics the F1 value [30] is used in thepaper for synthesis and the formula is
precision tr
tr + fr
recall tr
tr + fnr
F1 2 middot precision middot recallprecision + recall
(9)
4 Experiment and Analysis
41 Experimental Data
411 Data 1 e data in this paper are sorted into threetypes of enterprise human resource data including per-sonnel information personnel evaluation matrix values andpersonnel capability matrix values e datum was collectedfrom 4560 employees and 1233 positions with a sample sizeof 134540 [31]
412 Data 2 We included a large power supply company ina certain region the details of the company are as followscompany A is the main power supply and management unitin the region and the power supply covers 29 towns and 2forest farms under the jurisdiction of the region with a totalof 31 secondary power supply companies the company hasbeen established for more than 30 years and after a longperiod of development the companyrsquos staff structure andbusiness scope (service area) and quantity (mainly electricitysales and equipment) have changed dramatically By 2020more than 97 of the companyrsquos existing substations will beunmanned and the service will be based on a network ofeight 500 kV substations With the deepening of market-oriented reform of the national power grid the developmentof the company is facing new challenges with managementmethods needing improvement personnel structure need-ing optimization and power supply equipment and facilitiesneeding further upgrading In this context Company A hasput forward the human resources slogan of ldquopositive changetalent firstrdquo and strives to build a human resources man-agement system to adapt to the new situation striving tobecome a first-class power supply company in China and aleading player in the industry Information on the com-panyrsquos human resources and company performance ismainly summarized through the companyrsquos informationrelease [32]
Figure 4 depicts the trends of the number of employeesand electricity sales of company A from 2010 to 2020 It canbe seen that the number of employees of company A hasbeen maintaining a growth trend with a relatively slowgrowth rate and the electricity sales of company A except
for a certain downward trend around 2015 have generallymaintained a good growth and still maintained a goodgrowth in the recent year of total growth for the decline inelectricity demand during 2015 may be caused by factorssuch as the weakness of the domestic economy
Although the number of employees of the company hasmaintained growth the adjustment of relevant nationalpolicies and changes in the companyrsquos internal human re-sources structure have led to certain problems in the com-panyrsquos personnel structure which is manifested in the highdemand for power dispatching and transmission and sub-station personnel in the main power transmission network Inthis regard it is necessary for the company to plan its humanresources department in advance complete the human re-sources demand forecast as early as possible adjust andoptimize the personnel structure clean up the surplus per-sonnel and introduce the insufficient number of professionalsto serve the companyrsquos long-term development goals
rough the above basic situation it can be seen that theinformation related to the development history of companyA is relatively detailed and rich in data and the collected datashows certain volatility which is suitable for the analysis andprediction of its human resource demand by using double-loop neural network
42 Data 1 Experimental Test and Analysis of Results In thispaper the samples are divided into a training sample set anda test sample set e pseudocode of the neural networktesting procedure is given in Algorithm 1
en the feasibility of the algorithm in the paper iscompared with the experiments using the algorithm in thepaper CNN and the traditional statistical method for modeltraining and experimental analysis and the analysis evalu-ation indexes are accuracy recall and F1 value e ex-perimental results are shown in Table 1
times104times104
Electricity salesTotal personnel
0
1
2
3
4
5
6
2
22
24
26
28
3
32
34
36
38
4
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2010
Figure 4 Trend of the number of employees and electricity sales ofcompany A from 2010 to 2020
Computational Intelligence and Neuroscience 5
As can be seen from Table 1 the F1 value of the algo-rithm in the paper is significantly improved compared withthe other two algorithms and the traditional statisticalmethod is the least effective F1 value is 0678 in the case oflarge amount of data and the F1 value of common CNN isimproved to 0766 the best performance of the algorithm inthe paper is 0823 is indicates that the direct use of re-current convolutional network does not really improve thetraining features of the data while the method of usingglobal network plus local network in the paper can effectivelyimprove the hidden layer features of the data and thenimprove the data training quality and improve the matchingdegree and recommendation accuracy of the algorithm
43 Model Analytical Indexes and Screening in Data 2Experiments In this paper we established the analysis indexsystem of enterprise A and screened all the indexes based onthe gray correlation method to determine the final keyanalytical indexes of human resources demand
Based on the above principles the following analysisindexes were selected as the initial analysis system of HRdemand forecast for company A e specific sinks areshown in Table 2
After the construction of the original demand fore-casting analysis system we need to screen the indicators thataffect the human resource demand of company A ie thekey indicators e screening method is the gray correlationanalysis method of calculating the comprehensive correla-tion degree and the analysis results are summarized inTable 3 e correlation between each indicator and humanresources is calculated by the gray systemmodeling softwareand the specific results are as follows
ρa1 07896
ρa2 06157
ρa3 06875
ρa4 05792
ρa5 05680
ρa6 06080
(10)
e higher the value of the correlation the greater theinfluence on human resources demand By ranking the abovecorrelations we can get a1 a2 a3 a4 four variables withrelatively large values ie the length of transmission network
Input feature DOutput hybrid recurrent neural modelInitialize the hypermastigote which include the number of iterations t the learning rate L the hypermastigote of the recurrent neuralnetwork β1 β2 and the computational gradient of the model gt(1) i cycles from 1 to t(2) j cycles from 1 to t(3) Calculate the eigenvalues of each channel and substitute them into the function f(4) If j t then terminate the loop and execute the step 1(5) If jlt t go back to step 1(6) Extracting convolutional features to obtain F(7) Combine the F1 values and local model eigenvalues to obtain the probability values(8) Get the current job match value(9) Sort and output the final result(10) If ilt t then return to step (2) and loop through the i process(11) If i t end
ALGORITHM1 e neural network testing procedure
Table 1 Comparative experimental result
Model Accuracy Recall F1Model in this paper 08122 0832 0823Statistical model 0702 0649 0678Cyclic neural network 0755 0775 0766
Table 2 Company Arsquos human resource demand forecasts analysis system
Category 1 analysis index Category 2 analysis indicators Specific analysis indicators
Company development resources Core resources owned by the company Transmission network length aNumber of substations under jurisdiction a
Market development of the company Company market size Total number of users aTotal population served
Company development objectives Company size Annual revenue of the companyOperating conditions of the company Online electricity sales
6 Computational Intelligence and Neuroscience
the number of substations the total number of customers andthe amount of feed-in tariffs have a greater influence on theanalysis of human resources demand so the above fouranalysis indicators are selected as key indicators [33]
44Data 2Forecast e implementation of the double-loopneural network model consists of two stages first the keyindex prediction values are input into model to obtain thefinal prediction values
e predicted values of key analytical indicators ofhuman resource requirements of company A 2017ndash2019 areshown in Table 4
Figure 5 shows the error trend of the neural networktraining It can be seen that the output values obtained fromthe network training do not differ much from the optimaloutput valuese results are shown in Figure 6 R 09999 isobtained which initially indicates that our model is welltrained
Table 3 Summary of the results of the analysis of key indicators of human resources needs of enterprise A
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
3 Algorithm Flow
As mentioned in Section 22 the overall algorithm used isthe job matching recommendation algorithm e processdesign of the recommendation algorithm should also betailored to different application scenarios and data char-acteristics Currently used algorithms in the HR field areusually statistical algorithms that do not take into accountthe hidden features of the data and rely on simple scoringand expert judgment mechanisms which can easily lead toinformation scarcity problems [28]
In this paper we combine recurrent CNN and tradi-tional human resource allocation algorithm to design a jobmatching recommendation algorithm for human resourcefield e algorithm not only improves the problem of lowdata training quality of traditional algorithms but also ef-fectively improves the data computation efficiency by usingrecurrent neural networks [29] e core idea of the algo-rithm is to first extract the original features from the datawhich are consistent with the features required by traditionalhuman resources including the personnel evaluation matrixand the personnelrsquos ability matrixe data are extracted andencoded in an encoder and the encoded features are then fedas data input to the data input layer of RNN e data areprocessed using the recurrent convolutional layer to obtainthe job matching results and the algorithm flow is shown inFigure 3
First the data is collected and selected using a distributedstreaming collection method and grouped into a personnelevaluation matrix and a personnel competency matrix edata are then abstracted preprocessed encoded using anencoder and saved to a data warehouse and further enhancedusing a feature enhancement algorithm and fed into a RNN
e final job matching score is output and the HR rec-ommendation process is completed
e steps in the algorithm flow are described as follows
(1) Data collection using the distributed streaming datacollection method the format of manpower datavaries from company to company ereforemanpower data must be processed in a uniformformat including data rounding and conversionoperations
(2) Preprocessing of the raw data the datum is groupedso that a more comprehensive understanding of theHR model characteristics can be obtained e da-tum is also saved to the data warehouse so thatsubsequent data model training can be supported
(3) Perform feature enhancement the datum is obtainedfrom the data warehouse and the results of datagrouping are learned Fused data results are used asneural network data input for network training
(4) Recommendation result output the job matchingresults are ranked and then the reasonable job as-signment is made with reference to the score
31 Algorithm Evaluation Metrics e algorithms areevaluated by a certain number of metrics and the accurate
Input feature value
Channel 1 Channel n
Data splicing layer
Convolution Local connection
Averaging
Recommended results
Figure 2 Double-loop network model
Data acquisition
Capability evaluation
Data preprocessing
Data warehouse
Extract feature information
Enhanced feature model
Cyclic neural network
Personnel evaluation
User information
Person post matching score
Output results
Statistical information
Construct scoring matrix
Start
Figure 3 Algorithm flow
4 Computational Intelligence and Neuroscience
evaluation metrics are used in the evaluation of humanresource recommendation algorithms in this paper [13 22]
In the evaluation system the higher the accuracy andrecall the better the algorithm However in some scenariosaccuracy and recall may be contradictory so in order tosynthesize these two metrics the F1 value [30] is used in thepaper for synthesis and the formula is
precision tr
tr + fr
recall tr
tr + fnr
F1 2 middot precision middot recallprecision + recall
(9)
4 Experiment and Analysis
41 Experimental Data
411 Data 1 e data in this paper are sorted into threetypes of enterprise human resource data including per-sonnel information personnel evaluation matrix values andpersonnel capability matrix values e datum was collectedfrom 4560 employees and 1233 positions with a sample sizeof 134540 [31]
412 Data 2 We included a large power supply company ina certain region the details of the company are as followscompany A is the main power supply and management unitin the region and the power supply covers 29 towns and 2forest farms under the jurisdiction of the region with a totalof 31 secondary power supply companies the company hasbeen established for more than 30 years and after a longperiod of development the companyrsquos staff structure andbusiness scope (service area) and quantity (mainly electricitysales and equipment) have changed dramatically By 2020more than 97 of the companyrsquos existing substations will beunmanned and the service will be based on a network ofeight 500 kV substations With the deepening of market-oriented reform of the national power grid the developmentof the company is facing new challenges with managementmethods needing improvement personnel structure need-ing optimization and power supply equipment and facilitiesneeding further upgrading In this context Company A hasput forward the human resources slogan of ldquopositive changetalent firstrdquo and strives to build a human resources man-agement system to adapt to the new situation striving tobecome a first-class power supply company in China and aleading player in the industry Information on the com-panyrsquos human resources and company performance ismainly summarized through the companyrsquos informationrelease [32]
Figure 4 depicts the trends of the number of employeesand electricity sales of company A from 2010 to 2020 It canbe seen that the number of employees of company A hasbeen maintaining a growth trend with a relatively slowgrowth rate and the electricity sales of company A except
for a certain downward trend around 2015 have generallymaintained a good growth and still maintained a goodgrowth in the recent year of total growth for the decline inelectricity demand during 2015 may be caused by factorssuch as the weakness of the domestic economy
Although the number of employees of the company hasmaintained growth the adjustment of relevant nationalpolicies and changes in the companyrsquos internal human re-sources structure have led to certain problems in the com-panyrsquos personnel structure which is manifested in the highdemand for power dispatching and transmission and sub-station personnel in the main power transmission network Inthis regard it is necessary for the company to plan its humanresources department in advance complete the human re-sources demand forecast as early as possible adjust andoptimize the personnel structure clean up the surplus per-sonnel and introduce the insufficient number of professionalsto serve the companyrsquos long-term development goals
rough the above basic situation it can be seen that theinformation related to the development history of companyA is relatively detailed and rich in data and the collected datashows certain volatility which is suitable for the analysis andprediction of its human resource demand by using double-loop neural network
42 Data 1 Experimental Test and Analysis of Results In thispaper the samples are divided into a training sample set anda test sample set e pseudocode of the neural networktesting procedure is given in Algorithm 1
en the feasibility of the algorithm in the paper iscompared with the experiments using the algorithm in thepaper CNN and the traditional statistical method for modeltraining and experimental analysis and the analysis evalu-ation indexes are accuracy recall and F1 value e ex-perimental results are shown in Table 1
times104times104
Electricity salesTotal personnel
0
1
2
3
4
5
6
2
22
24
26
28
3
32
34
36
38
4
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2010
Figure 4 Trend of the number of employees and electricity sales ofcompany A from 2010 to 2020
Computational Intelligence and Neuroscience 5
As can be seen from Table 1 the F1 value of the algo-rithm in the paper is significantly improved compared withthe other two algorithms and the traditional statisticalmethod is the least effective F1 value is 0678 in the case oflarge amount of data and the F1 value of common CNN isimproved to 0766 the best performance of the algorithm inthe paper is 0823 is indicates that the direct use of re-current convolutional network does not really improve thetraining features of the data while the method of usingglobal network plus local network in the paper can effectivelyimprove the hidden layer features of the data and thenimprove the data training quality and improve the matchingdegree and recommendation accuracy of the algorithm
43 Model Analytical Indexes and Screening in Data 2Experiments In this paper we established the analysis indexsystem of enterprise A and screened all the indexes based onthe gray correlation method to determine the final keyanalytical indexes of human resources demand
Based on the above principles the following analysisindexes were selected as the initial analysis system of HRdemand forecast for company A e specific sinks areshown in Table 2
After the construction of the original demand fore-casting analysis system we need to screen the indicators thataffect the human resource demand of company A ie thekey indicators e screening method is the gray correlationanalysis method of calculating the comprehensive correla-tion degree and the analysis results are summarized inTable 3 e correlation between each indicator and humanresources is calculated by the gray systemmodeling softwareand the specific results are as follows
ρa1 07896
ρa2 06157
ρa3 06875
ρa4 05792
ρa5 05680
ρa6 06080
(10)
e higher the value of the correlation the greater theinfluence on human resources demand By ranking the abovecorrelations we can get a1 a2 a3 a4 four variables withrelatively large values ie the length of transmission network
Input feature DOutput hybrid recurrent neural modelInitialize the hypermastigote which include the number of iterations t the learning rate L the hypermastigote of the recurrent neuralnetwork β1 β2 and the computational gradient of the model gt(1) i cycles from 1 to t(2) j cycles from 1 to t(3) Calculate the eigenvalues of each channel and substitute them into the function f(4) If j t then terminate the loop and execute the step 1(5) If jlt t go back to step 1(6) Extracting convolutional features to obtain F(7) Combine the F1 values and local model eigenvalues to obtain the probability values(8) Get the current job match value(9) Sort and output the final result(10) If ilt t then return to step (2) and loop through the i process(11) If i t end
ALGORITHM1 e neural network testing procedure
Table 1 Comparative experimental result
Model Accuracy Recall F1Model in this paper 08122 0832 0823Statistical model 0702 0649 0678Cyclic neural network 0755 0775 0766
Table 2 Company Arsquos human resource demand forecasts analysis system
Category 1 analysis index Category 2 analysis indicators Specific analysis indicators
Company development resources Core resources owned by the company Transmission network length aNumber of substations under jurisdiction a
Market development of the company Company market size Total number of users aTotal population served
Company development objectives Company size Annual revenue of the companyOperating conditions of the company Online electricity sales
6 Computational Intelligence and Neuroscience
the number of substations the total number of customers andthe amount of feed-in tariffs have a greater influence on theanalysis of human resources demand so the above fouranalysis indicators are selected as key indicators [33]
44Data 2Forecast e implementation of the double-loopneural network model consists of two stages first the keyindex prediction values are input into model to obtain thefinal prediction values
e predicted values of key analytical indicators ofhuman resource requirements of company A 2017ndash2019 areshown in Table 4
Figure 5 shows the error trend of the neural networktraining It can be seen that the output values obtained fromthe network training do not differ much from the optimaloutput valuese results are shown in Figure 6 R 09999 isobtained which initially indicates that our model is welltrained
Table 3 Summary of the results of the analysis of key indicators of human resources needs of enterprise A
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
evaluation metrics are used in the evaluation of humanresource recommendation algorithms in this paper [13 22]
In the evaluation system the higher the accuracy andrecall the better the algorithm However in some scenariosaccuracy and recall may be contradictory so in order tosynthesize these two metrics the F1 value [30] is used in thepaper for synthesis and the formula is
precision tr
tr + fr
recall tr
tr + fnr
F1 2 middot precision middot recallprecision + recall
(9)
4 Experiment and Analysis
41 Experimental Data
411 Data 1 e data in this paper are sorted into threetypes of enterprise human resource data including per-sonnel information personnel evaluation matrix values andpersonnel capability matrix values e datum was collectedfrom 4560 employees and 1233 positions with a sample sizeof 134540 [31]
412 Data 2 We included a large power supply company ina certain region the details of the company are as followscompany A is the main power supply and management unitin the region and the power supply covers 29 towns and 2forest farms under the jurisdiction of the region with a totalof 31 secondary power supply companies the company hasbeen established for more than 30 years and after a longperiod of development the companyrsquos staff structure andbusiness scope (service area) and quantity (mainly electricitysales and equipment) have changed dramatically By 2020more than 97 of the companyrsquos existing substations will beunmanned and the service will be based on a network ofeight 500 kV substations With the deepening of market-oriented reform of the national power grid the developmentof the company is facing new challenges with managementmethods needing improvement personnel structure need-ing optimization and power supply equipment and facilitiesneeding further upgrading In this context Company A hasput forward the human resources slogan of ldquopositive changetalent firstrdquo and strives to build a human resources man-agement system to adapt to the new situation striving tobecome a first-class power supply company in China and aleading player in the industry Information on the com-panyrsquos human resources and company performance ismainly summarized through the companyrsquos informationrelease [32]
Figure 4 depicts the trends of the number of employeesand electricity sales of company A from 2010 to 2020 It canbe seen that the number of employees of company A hasbeen maintaining a growth trend with a relatively slowgrowth rate and the electricity sales of company A except
for a certain downward trend around 2015 have generallymaintained a good growth and still maintained a goodgrowth in the recent year of total growth for the decline inelectricity demand during 2015 may be caused by factorssuch as the weakness of the domestic economy
Although the number of employees of the company hasmaintained growth the adjustment of relevant nationalpolicies and changes in the companyrsquos internal human re-sources structure have led to certain problems in the com-panyrsquos personnel structure which is manifested in the highdemand for power dispatching and transmission and sub-station personnel in the main power transmission network Inthis regard it is necessary for the company to plan its humanresources department in advance complete the human re-sources demand forecast as early as possible adjust andoptimize the personnel structure clean up the surplus per-sonnel and introduce the insufficient number of professionalsto serve the companyrsquos long-term development goals
rough the above basic situation it can be seen that theinformation related to the development history of companyA is relatively detailed and rich in data and the collected datashows certain volatility which is suitable for the analysis andprediction of its human resource demand by using double-loop neural network
42 Data 1 Experimental Test and Analysis of Results In thispaper the samples are divided into a training sample set anda test sample set e pseudocode of the neural networktesting procedure is given in Algorithm 1
en the feasibility of the algorithm in the paper iscompared with the experiments using the algorithm in thepaper CNN and the traditional statistical method for modeltraining and experimental analysis and the analysis evalu-ation indexes are accuracy recall and F1 value e ex-perimental results are shown in Table 1
times104times104
Electricity salesTotal personnel
0
1
2
3
4
5
6
2
22
24
26
28
3
32
34
36
38
4
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2010
Figure 4 Trend of the number of employees and electricity sales ofcompany A from 2010 to 2020
Computational Intelligence and Neuroscience 5
As can be seen from Table 1 the F1 value of the algo-rithm in the paper is significantly improved compared withthe other two algorithms and the traditional statisticalmethod is the least effective F1 value is 0678 in the case oflarge amount of data and the F1 value of common CNN isimproved to 0766 the best performance of the algorithm inthe paper is 0823 is indicates that the direct use of re-current convolutional network does not really improve thetraining features of the data while the method of usingglobal network plus local network in the paper can effectivelyimprove the hidden layer features of the data and thenimprove the data training quality and improve the matchingdegree and recommendation accuracy of the algorithm
43 Model Analytical Indexes and Screening in Data 2Experiments In this paper we established the analysis indexsystem of enterprise A and screened all the indexes based onthe gray correlation method to determine the final keyanalytical indexes of human resources demand
Based on the above principles the following analysisindexes were selected as the initial analysis system of HRdemand forecast for company A e specific sinks areshown in Table 2
After the construction of the original demand fore-casting analysis system we need to screen the indicators thataffect the human resource demand of company A ie thekey indicators e screening method is the gray correlationanalysis method of calculating the comprehensive correla-tion degree and the analysis results are summarized inTable 3 e correlation between each indicator and humanresources is calculated by the gray systemmodeling softwareand the specific results are as follows
ρa1 07896
ρa2 06157
ρa3 06875
ρa4 05792
ρa5 05680
ρa6 06080
(10)
e higher the value of the correlation the greater theinfluence on human resources demand By ranking the abovecorrelations we can get a1 a2 a3 a4 four variables withrelatively large values ie the length of transmission network
Input feature DOutput hybrid recurrent neural modelInitialize the hypermastigote which include the number of iterations t the learning rate L the hypermastigote of the recurrent neuralnetwork β1 β2 and the computational gradient of the model gt(1) i cycles from 1 to t(2) j cycles from 1 to t(3) Calculate the eigenvalues of each channel and substitute them into the function f(4) If j t then terminate the loop and execute the step 1(5) If jlt t go back to step 1(6) Extracting convolutional features to obtain F(7) Combine the F1 values and local model eigenvalues to obtain the probability values(8) Get the current job match value(9) Sort and output the final result(10) If ilt t then return to step (2) and loop through the i process(11) If i t end
ALGORITHM1 e neural network testing procedure
Table 1 Comparative experimental result
Model Accuracy Recall F1Model in this paper 08122 0832 0823Statistical model 0702 0649 0678Cyclic neural network 0755 0775 0766
Table 2 Company Arsquos human resource demand forecasts analysis system
Category 1 analysis index Category 2 analysis indicators Specific analysis indicators
Company development resources Core resources owned by the company Transmission network length aNumber of substations under jurisdiction a
Market development of the company Company market size Total number of users aTotal population served
Company development objectives Company size Annual revenue of the companyOperating conditions of the company Online electricity sales
6 Computational Intelligence and Neuroscience
the number of substations the total number of customers andthe amount of feed-in tariffs have a greater influence on theanalysis of human resources demand so the above fouranalysis indicators are selected as key indicators [33]
44Data 2Forecast e implementation of the double-loopneural network model consists of two stages first the keyindex prediction values are input into model to obtain thefinal prediction values
e predicted values of key analytical indicators ofhuman resource requirements of company A 2017ndash2019 areshown in Table 4
Figure 5 shows the error trend of the neural networktraining It can be seen that the output values obtained fromthe network training do not differ much from the optimaloutput valuese results are shown in Figure 6 R 09999 isobtained which initially indicates that our model is welltrained
Table 3 Summary of the results of the analysis of key indicators of human resources needs of enterprise A
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
As can be seen from Table 1 the F1 value of the algo-rithm in the paper is significantly improved compared withthe other two algorithms and the traditional statisticalmethod is the least effective F1 value is 0678 in the case oflarge amount of data and the F1 value of common CNN isimproved to 0766 the best performance of the algorithm inthe paper is 0823 is indicates that the direct use of re-current convolutional network does not really improve thetraining features of the data while the method of usingglobal network plus local network in the paper can effectivelyimprove the hidden layer features of the data and thenimprove the data training quality and improve the matchingdegree and recommendation accuracy of the algorithm
43 Model Analytical Indexes and Screening in Data 2Experiments In this paper we established the analysis indexsystem of enterprise A and screened all the indexes based onthe gray correlation method to determine the final keyanalytical indexes of human resources demand
Based on the above principles the following analysisindexes were selected as the initial analysis system of HRdemand forecast for company A e specific sinks areshown in Table 2
After the construction of the original demand fore-casting analysis system we need to screen the indicators thataffect the human resource demand of company A ie thekey indicators e screening method is the gray correlationanalysis method of calculating the comprehensive correla-tion degree and the analysis results are summarized inTable 3 e correlation between each indicator and humanresources is calculated by the gray systemmodeling softwareand the specific results are as follows
ρa1 07896
ρa2 06157
ρa3 06875
ρa4 05792
ρa5 05680
ρa6 06080
(10)
e higher the value of the correlation the greater theinfluence on human resources demand By ranking the abovecorrelations we can get a1 a2 a3 a4 four variables withrelatively large values ie the length of transmission network
Input feature DOutput hybrid recurrent neural modelInitialize the hypermastigote which include the number of iterations t the learning rate L the hypermastigote of the recurrent neuralnetwork β1 β2 and the computational gradient of the model gt(1) i cycles from 1 to t(2) j cycles from 1 to t(3) Calculate the eigenvalues of each channel and substitute them into the function f(4) If j t then terminate the loop and execute the step 1(5) If jlt t go back to step 1(6) Extracting convolutional features to obtain F(7) Combine the F1 values and local model eigenvalues to obtain the probability values(8) Get the current job match value(9) Sort and output the final result(10) If ilt t then return to step (2) and loop through the i process(11) If i t end
ALGORITHM1 e neural network testing procedure
Table 1 Comparative experimental result
Model Accuracy Recall F1Model in this paper 08122 0832 0823Statistical model 0702 0649 0678Cyclic neural network 0755 0775 0766
Table 2 Company Arsquos human resource demand forecasts analysis system
Category 1 analysis index Category 2 analysis indicators Specific analysis indicators
Company development resources Core resources owned by the company Transmission network length aNumber of substations under jurisdiction a
Market development of the company Company market size Total number of users aTotal population served
Company development objectives Company size Annual revenue of the companyOperating conditions of the company Online electricity sales
6 Computational Intelligence and Neuroscience
the number of substations the total number of customers andthe amount of feed-in tariffs have a greater influence on theanalysis of human resources demand so the above fouranalysis indicators are selected as key indicators [33]
44Data 2Forecast e implementation of the double-loopneural network model consists of two stages first the keyindex prediction values are input into model to obtain thefinal prediction values
e predicted values of key analytical indicators ofhuman resource requirements of company A 2017ndash2019 areshown in Table 4
Figure 5 shows the error trend of the neural networktraining It can be seen that the output values obtained fromthe network training do not differ much from the optimaloutput valuese results are shown in Figure 6 R 09999 isobtained which initially indicates that our model is welltrained
Table 3 Summary of the results of the analysis of key indicators of human resources needs of enterprise A
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
the number of substations the total number of customers andthe amount of feed-in tariffs have a greater influence on theanalysis of human resources demand so the above fouranalysis indicators are selected as key indicators [33]
44Data 2Forecast e implementation of the double-loopneural network model consists of two stages first the keyindex prediction values are input into model to obtain thefinal prediction values
e predicted values of key analytical indicators ofhuman resource requirements of company A 2017ndash2019 areshown in Table 4
Figure 5 shows the error trend of the neural networktraining It can be seen that the output values obtained fromthe network training do not differ much from the optimaloutput valuese results are shown in Figure 6 R 09999 isobtained which initially indicates that our model is welltrained
Table 3 Summary of the results of the analysis of key indicators of human resources needs of enterprise A
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
emodel was simulated using the historical data of 2019and the actual output value of 6437 was obtained for 2019 andthe relative error of its calculation was 1067 which was lessthan 3 e data obtained from the GM (1 1) model wasused as the input of the neural network to predict the humanresource demand of enterprise A from 2017 to 2019 and therelevant results are summarized as shown in Table 5
According to the forecast results enterprise A needs tofurther increase the total number of staff in T e specificstaffing requirements need to be adjusted according to theactual situation of the company
45 Comparison of Different Models In order to visuallycompare the prediction effects of different model the pre-diction analysis of the two was carried out separately in thispaper and the specific results are shown in Table 6
From the HR demand forecasting values of the twomodels in Table 6 it is clear that the forecasting results of thegray BP network forecasting model for each year are closerto the true values than the GM (1 1) model [17 34] whichhas better forecasting accuracy meanwhile the averagerelative error of the dual recurrent neural network fore-casting model is only 01481 which indicates that themodel has very high forecasting accuracy is indirectlyindicates the applicability and reliability of the two-circu-lation neural network prediction model selected in thispaper
5 Conclusions
e essence of HR scheduling model is to analyze HR dataand calculate the job matching score en the scheduling ofpersonnel is performed based on the job matching scorewhich can be abstracted as a recommendation model inessence In the paper the basic neural network is improvedby using a combination of a double-loop neural networkmodel a global model and a local model and the datafeatures after the hierarchical operation of the model areused as the network output and then the data is processedBy using a hierarchical model structure for network con-struction we finally achieve high accuracy job matching andrecommendation
Data Availability
e dataset used in this paper is available from the corre-sponding author upon request
Conflicts of Interest
e authors declare that they have no conflicts of interestregarding this work
Acknowledgments
is work was supported by China Academy of Manage-ment Science Research on Practice Teaching System ofBusiness Administration Specialty in Colleges and Uni-versities Based on the Integration of Industry and Educationunder grant no ZJKY4990 and China Academy of Man-agement Science Research on Idea Construction of Inno-vation and Entrepreneurship Education in Colleges andUniversities and Reform of Talents Training Mode undergrant no GJY4422
Our methodCNN
02
03
04
05
06
07
08A
ccur
acy
20 40 60 80 100 120 1400Epoch
Figure 6 Prediction results of our model
Table 5 Summary of the prediction results of the trained BPnetwork
Particular year 2107 2018 2019Predicted value (person) 6483 6549 6699
Table 6 Summary of the prediction results of the gray predictionmodel and our model
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
References
[1] W Zhao S Pu and D Jiang ldquoA human resource allocationmethod for business processes using team fault linesrdquo AppliedIntelligence vol 50 no 5 pp 1ndash14 2020
[2] C Gao and H Sun ldquoStrategic transformation of human re-source management model of ocean engineering an ex-ploratory studyrdquo Journal of Coastal Research vol 106 no 1p 117 2020
[3] R J Burke P Spurgeon and C L Cooper ldquoe innovationimperative in health care organisations critical role of humanresource management in the cost quality and productivityequationrdquo Neurosurgery vol 77 no 1 pp 67ndash80 2015
[4] K E Mills D M Weary and M A G von KeyserlingkldquoGraduate student literature review challenges and oppor-tunities for human resource management on dairy farmsrdquoJournal of Dairy Science vol 104 no 1 pp 1192ndash1202 2021
[5] R Li Q Liu J Gui D Gu and H Hu ldquoIndoor relocalizationin challenging environments with dual-stream convolutionalneural networksrdquo IEEE Transactions on Automation Scienceand Engineering vol 15 no 2 pp 651ndash662 2017
[6] A Sareen ldquoP19Development of a guideline on managementof medicines for patients on a ketogenic dietrdquo Archives ofDisease in Childhood vol 103 no 2 p e2 2018
[7] M Osman ldquoStudy of effect of implementation of humanresource management profiling in small holder dairy farmsrdquoDairy Science amp Technology vol 1 no 3 pp 41ndash45 2015
[8] G Atluri A Karpatne and V Kumar ldquoSpatio-temporal dataminingrdquo ACM Computing Surveys vol 51 no 4 pp 1ndash412018
[9] O S Yee S Sagadevan and N H A H Malim ldquoCredit cardfraud detection using machine learning as data miningtechniquerdquo Journal of Telecommunication Electronic andComputer Engineering vol 10 no 1ndash4 pp 23ndash27 2018
[10] D McNeish D G Dumas and K J Grimm ldquoEstimating newquantities from longitudinal test scores to improve forecastsof future performancerdquo Multivariate Behavioral Researchvol 55 no 6 pp 894ndash909 2020
[11] D Tomasko M Alderson R Burnes et al ldquoWidespreadrecovery of seagrass coverage in Southwest Florida (USA)temporal and spatial trends and management actions re-sponsible for successrdquo Marine Pollution Bulletin vol 135pp 1128ndash1137 2018
[12] A Kamilaris and F X Prenafeta-Boldu ldquoDeep learning inagriculture a surveyrdquo Computers and Electronics in Agri-culture vol 147 pp 70ndash90 2018
[13] C J C Jabbour and A B L de Sousa Jabbour ldquoGreen humanresource management and green supply chain managementlinking two emerging agendasrdquo Journal of Cleaner Productionvol 112 no 3 pp 1824ndash1833 2016
[14] M Salah K Altalla A Salah and S S Abu-Naser ldquoPredictingmedical expenses using artificial neural networkrdquo Interna-tional Journal of Engineering and Information Systems vol 2no 20 pp 11ndash17 2018
[15] D Wu C Zhang L Ji R Ran H Wu and Y Xu ldquoForest firerecognition based on feature extraction from multi-viewimagesrdquo Traitement du Signal vol 38 no 3 pp 775ndash7832021
[16] H Boutmaghzoute and K Moustaghfir ldquoExploring the re-lationship between corporate social responsibility actions andemployee retention a human resource management per-spectiverdquo Human Systems Management vol 12 pp 1ndash132021
[17] D W Tjondronegoro and Y-P P Chen ldquoKnowledge-dis-counted event detection in sports videordquo IEEE Transactionson Systems Man and Cybernetics-Part A Systems andHumans vol 40 no 5 pp 1009ndash1024 2010
[18] Z Gao P Wang H Wang M Xu and W Li ldquoA review ofdynamic maps for 3D human motion recognition usingConvNets and its improvementrdquo Neural Processing Lettersvol 52 no 2 pp 1501ndash1515 2020
[19] X-B Fu S-L Yue and D-Y Pan ldquoCamera-based basketballscoring detection using convolutional neural networkrdquo In-ternational Journal of Automation and Computing vol 18no 2 pp 266ndash276 2021
[20] B Li and X Xu ldquoApplication of artificial intelligence inbasketball sportrdquo Journal of Education Health and Sportvol 11 no 7 pp 54ndash67 2021
[21] D Arena A C Tsolakis S Zikos et al ldquoHuman resourceoptimisation through semantically enriched datardquo Interna-tional Journal of Production Research vol 56 no 7-8pp 2855ndash2877 2018
[22] B-B Lee R Ibrahim S-Y Chu N A Zulkifli andP Ravindra ldquoAlginate liquid core capsule formation using thesimple extrusion dripping methodrdquo Journal of Polymer En-gineering vol 35 no 4 pp 311ndash318 2015
[23] N J Navimipour ldquoA formal approach for the specificationand verification of a trustworthy human resource discoverymechanism in the expert cloudrdquo Expert Systems with Ap-plications vol 42 no 15-16 pp 6112ndash6131 2015
[24] C Zhang T Xie K Yang et al ldquoPositioning optimisationbased on particle quality prediction in wireless sensor net-worksrdquo IET Networks vol 8 no 2 pp 107ndash113 2019
[25] N A Rahmad N A J Sufri N H Muzamil andM A AsrsquoarildquoBadminton player detection using faster region convolu-tional neural networkrdquo Indonesian Journal of Electrical En-gineering and Computer Science vol 14 no 3 pp 1330ndash13352019
[26] Y Wang X Ma Z Chen Y Luo J Yi and J BaileyldquoSymmetric cross entropy for robust learning with noisylabelsrdquo in Proceedings of the IEEECVF International Con-ference on Computer Vision pp 322ndash330 Seoul South Korea2019
[27] D Liang J Li and R Qu ldquoSensorless control of permanentmagnet synchronous machine based on second-order sliding-mode observer with online resistance estimationrdquo IEEETransactions on Industry Applications vol 53 no 4pp 3672ndash3682 2017
[28] B Mahaseni E R M Faizal and R G Raj ldquoSpotting footballevents using two-stream convolutional neural network anddilated recurrent neural networkrdquo IEEE Access vol 9pp 61929ndash61942 2021
[29] M Trinidad-Fernandez D Beckwee A Cuesta-Vargas et alldquoDifferences in movement limitations in different low backpain severity in functional tests using an RGB-D camerardquoJournal of Biomechanics vol 116 p 110212 2021
[30] T Xie C Zhang Z Zhang and K Yang ldquoUtilizing activesensor nodes in smart environments for optimal communi-cation coveragerdquo IEEE Access vol 7 pp 11338ndash11348 2018
[31] H Tetiana C Maryna and K Lidiia ldquoInnovative model ofenterprises personnel incentives evaluationrdquo Academy ofStrategic Management Journal vol 17 no 3 pp 1ndash6 2018
[32] N Ben Moussa and R El Arbi ldquoe impact of human re-sources information systems on individual innovation ca-pability in Tunisian companies the moderating role ofaffective commitmentrdquo European Research on Managementand Business Economics vol 26 no 1 pp 18ndash25 2020
Computational Intelligence and Neuroscience 9
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
10 Computational Intelligence and Neuroscience
[33] K Zhong Y Wang J Pei S Tang and Z Han ldquoSuper ef-ficiency SBM-DEA and neural network for performanceevaluationrdquo Information Processing amp Management vol 58no 6 p 102728 2021
[34] S Chen Y Liu L Wei and B Guan ldquoPS-FW a hybrid al-gorithm based on particle swarm and fireworks for globaloptimizationrdquo Computational Intelligence and Neurosciencevol 2018 Article ID 6094685 27 pages 2018
