of 19
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
1/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
Data Mining Analytics to Minimize Logistics Cost
Abraham Paul11School of Information
Technology and EngineeringVIT University
Vellore, TNIndia
V.Saravanan22Department of Computer
ApplicationsDr. NGP Institute of
TechnologyCoimbatore, TN
P. Ranjit Jeba Thangaiah33Department of Computer
ApplicationsKarunya University
Coimbatore, TNIndia
Abstract
Transportation and Logistics are a major sector of the economy; however data analysis in th is
domain has remained largely in the province of optimization. The potential of data mining and
knowledge discovery techni ques is largely untapped. Thi s paper is about solving vehicle routing
problem using Operation Research (OR) approach in analysis and design phases and we use JAVA
programming language to model the problem. The resul ts obtained fr om both solutions are compared
in order to make analysis and prove the object-ori ented model corr ectness. We proved that both
resul ts are identical and have the same resul ts when solving the problem using the fi ve methods:
northwest corner method; min imum cost method; r ow minimum cost method; column minimum cost
method, and Vogels approximation method.
We also suggest several chall enging problems to precipi tate research and galvani ze fu tur e work i n
thi s area.
Keywords: Data Mining, Transportation problem, Linear Programming, Object-orientedProgramming
1. Introduction
Data mining currently is a hot topic research area and is applied in database, artificial intelligence,statistics, and so on. It may discover valuable knowledge and the patterns in the large-scale database forusers.
Third Party Logistics (3PL) providers are competing worldwide through Logistics Optimization toreduce costs while achieving high delivery standards. Logistics optimization is currently the biggestopportunity for most companies in order to attain significant reduction in operational costs. It can save3PL providers up to 1040% on operational costs by improving decisions such as the optimal selectionof inventory placement and transportation modes.
Transportation and logistics are an important sector of the economy. Transportation consumes 60%of oil worldwide [9], and the number is increasing. Data mining has lead to significant gains in other
areas, and should also be used to improve this sector of our economy. Computer use is widespread intransportation and logistics. Inventory management, parcel tracking and even on-truck location] sensorsprovide a wealth of data. This seems a natural application area for data mining, however, to date, therehave been few success stories. There has been some mining with freight flow data, but with transactionalcharacteristics of freight and events such as safety/accident records rather than the geometry of thenetwork. For example, classification on safety/accident records might find that trucks are prone toaccidents at 7:00AM on east - west roads (i.e., when the sun is in the drivers eyes.) A similar problemcould be to find conditions in which trucks suffer from mechanical failure to predict a requirement for
March Issue Page 89 of 107 ISSN 2229 5216
mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]7/29/2019 Data Mining Analytics to Minimize Logistics Cost
2/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
maintenance. Existing methods that utilize the network structure fall within the domain of optimization.Optimization techniques have long been computerized, but do not provide the kinds of insights that arethe goal of data mining. These optimization techniques generally fix certain constraints and allow amanageable number of free variables to be adjusted to achieve minimal cost. The goal of our datamining is not to better optimize these free variables, but to generate knowledge that enables a business toadjust its process to modify the constraints, enabling the optimization to generate a lower cost solution.Optimization may realize a feasible solution, but data mining could find patterns where constraints couldbe modified or re-defined to provide a more robust optimal solution. One of the few reported datamining successes in this area involves management of inventory levels [1,2], where the constraints onsafety stocks were modified to achieve lower inventory and higher in stock rates.
1.1 Logistics optimization
The third-party logistics provider (3PL) performs logistics services on behalf of another company.There are two primary kinds of 3PL Providers. Type 1 assumes infinite production on supply depot andtheir only concern is to meet the demand at the delivery nodes. Type 2 3PL providers perform thefunctions of meeting the demands and also taking into consideration the safety stock and the economicorder quantity at the demand node. In order to achieve this, these companies need to maintain certaininventory at their warehouse in order to perform on- time delivery and achieve a certain service level.
Figure 1: A Common Network
Logistics optimization problem for the Type 1 3PL can be solved with the Vehicle Routing Problem inorder to minimize transportation costs. For the Type 2 3PL Company, logistics optimization is achievedby vehicle routing problem and additionally minimizing the inventory carrying costs at their warehouse.A logistics optimization problem starts with defining the network. Figure 1 illustrates a commonnetwork with one supply depot and various demand nodes. The Vehicle Routing Problem determines theroutes for the network. Vehicle Routing Problem for all 3PL providers The Vehicle Routing Problemdetermines the set of routes with overall minimum route cost which services all the demands for a givena fleet of vehicles, a service depot, and several demand nodes. Figure 2 shows different types of vehicleroutings and also lists the solutions to find the optimized routes.
Figure 2: The Vehicle Routing Problem
March Issue Page 90 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
3/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
1.2 Inventory optimization for type 2 3PL providers
For the second type of 3PL, the additional objective is to meet the desired service level and supplyingthe Economic Order Quantity (EOQ) at the demand nodes. This is done cost effectively by maintaininginventories at the 3PL provider warehouse. This can be achieved by the inventory optimization modelwith the following objective function:
Minimize inventory holding cost, shortage and backlogs.The model minimizes the inventory holding costs ensuring the best feasible solution considering allthe constraints. The optimization model considers the following variable and constraints: Demands and EOQ calculations at the delivery nodes Products and/or product groups Penalties on safety stock (Backlog penalties, Shortage penalties) Warehouse Constraints (Warehouse capacity) Inventory Constraints (Inventory holding capacity by product and/or product group) Inventory holding costs
The output of the model provides the optimum inventory levels by product at the warehouse in order tomeet the demands at a certain service level and also minimize the inventory holding cost.
1.3 Other optimization models
Additional function of consolidation of inventories can be added in the above described model, basedon the requirement. Here the objective is to convert all Less than Truck Loads (LTL) to Full TruckLoads (FTL). Also, an optimum daily dispatch plan can be created by for a given network and for arange of product. The objective function is to minimize transportation costs, inventory holding costs andpenalty costs in order to minimize consumption of safety stock and avoid backlogs.The function can be modeled mathematically as below:
Minimize Transportation Costs) + (Inventory Holding Costs) + (Penalties)The objective function is subjected to the following parameters: Demand shortage Net demand Actual opening inventory Safety stock shortage Supply Penalty cost for late shipment,early shipment and safety
stock consumption In transit supply Shipment quantity
Node capacity violation Transportation Costs (Fixed and Variable) Product quantity No. of Transports Inventory carrying cost
Logistics optimization provides strength and value to the 3PL supply chain. Companies are allocatingresources and money to make their logistics departments more efficient and economic. Through properplanning and using the right logistics optimization tools, companies can acquire considerable costreduction across their logistics & distribution operation and can accomplish better delivery standards.
1.4 Predictive analytics
We use historical data intelligently to develop a view of future market trends and help our clients focuson the right audiences thereby developing their competitive edges.
2. Data mining method
The analytical techniques used in data mining are often well-known mathematical algorithms andtechniques. What is new is the application of those techniques to general business problems madepossible by the increased availability of data and inexpensive storage and processing power. Also, theuse of graphical interfaces has led to tools becoming available that business experts can easily use. Someof the tools used for data mining are [2]:
March Issue Page 91 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
4/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
1) Artificial neural networks. On-linear predictive models that learn through training and resemblebiological neural networks in structure.2) Decision trees. Tree-shaped structures that represent sets of decisions. These decisions generate rulesfor the classification of a dataset.3) Rule induction. The extraction of useful if-then rules from data based on statistical significance.4) Genetic algorithms. Optimization techniques based on the concepts of genetic combination, mutation,and natural selection.5) Nearest neighbor. A classification technique that classifies each record based on the records mostsimilar to it in an historical database.
2.1. Association rules mining
Association rule mining finds interesting associations and/or correlation relationships among large setof data items. Association rules show attributes value conditions that occur frequently together in a givendataset. Association rules identify collections of data attributes that are statistically related in theunderlying data. An association rule is of the form X => Y where X and Y are disjoint conjunctions ofattribute-value pairs. The confidence of the rule is the conditional probability of Y given X, Pr(Y|X), andthe support of the rule is the prior probability of X and Y, Pr(X and Y). Here probability is taken to bethe observed frequency in the data set. The traditional association rule mining problem can be describedas follows. Given a database of transactions, a minimal confidence threshold and a minimal support
threshold, find all association rules whose confidence and support are above the correspondingthresholds. The steps of association rules mining based on data cube as followsStep1: mining frequent item-set which satisfies the minimum support on data cubeStep2: association rules of frequent item-set are generated.
2.2. Intelligent transportation systems application
In the ITS, violating regulation information, driver information and vehicle information are quitetedious. This original data is difficult to mine out the effective patterns. Data selection and cleaning arefirst. Then integrate the data, process data and start mining. Finally, the patterns that are mined out areevaluated. Mining process is as shown in Fig.1
Figure 3: Mining process
3. Transportation problem
The first main purpose is solving transportation problem using five methods of transportation modelby linear programming (LP). The second main purpose is solving transportation problem by object-oriented programming. C++ programming language is used to get the solution. The results obtain fromboth LP and object oriented programming solutions are compared.The five methods for solving Transportation problem are:
March Issue Page 92 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
5/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
1. Northwest Corner method2. Minimum cost method3. Vogels approximation method4. Row Minimum Method5. Column Minimum Method
This paper introduces methods for solving transportation problem by C++ programming language; weuse flow chart, algorithms and consider the importance of defining a problem sufficiently and whatassumptions we may consider during the solution. Solving transportation problem by computer involvesserves of steps: define the problem, analysis the problem and formulate a method to solve it, describe thesolution in the form of an algorithm, draw a flow chart of the algorithm, write the computer program,compile and run the program, test the program and interpretation of results.
We design Object-Oriented Model as decision support tool to evaluate the solution for the fivemethods using JAVA language. After designing the five models (the five programs) we comparebetween each solution using JAVA programs and LP solution which have the same result. Comparisonbetween different solutions is done for choosing less value of the objective function so that the user willbe able to make decision.
4. Transportation model
Transportation model is a special type of networks problems that for shipping a commodity fromsource (e.g. factories) to destinations (e.g. warehouse). Transportation model deal with get theminimum-cost plan to transport a commodity from a number of sources(m) to number of destination (n).
Let si is the number of supply units required at source i (i=1, 2, 3. m), dj is the number of demandunits required at destination j (j=1, 2, 3.. n) and c ij represent the unit transportation cost fortransporting the units from sources i to destination j.
Using linear programming method to solve transportation problem, we determine the value ofobjective function which minimize the cost for transporting and also determine the number of unit canbe transported from source i to destination j. If xij is number of units shipped from source i to destinationj. the equivalent linear programming model[7] will be as follows.The objective function
Subject to
Figure 4: Network representation of the transportation problem
for i=1,2,..m.for j=1,2,...,n and xij 0 for all i to j.
March Issue Page 93 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
6/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
A transportation problem is said to be balanced if the total supply from all sources equals the totaldemand in all destinations.
5. METHODS FOR SOLVING TRANSPORTATION PROBLEM
There are five methods to determine the solution for balanced transportation problem:The five methods differ in the "quality" of the starting basic solution they produce and better startingsolution yields a smaller objective value. We present the five methods and an illustrative example issolved by these five methods.
5.1 North West-Corner Method
The method starts at the northwest-corner cell (route) of the tableau (variable x11)(i)Allocate as much as possible to the selected cell and adjust the associated a mounts of supply anddemand by subtracting the allocated amount.(ii)Cross out the row or Column with zero supply or demand to indicate that no further assignments canbe made in that row or column. If both a row and a column net to zero simultaneously, cross out oneonly and leave a zero supply (demand in the uncrossed-out row column).(iii) If exactly one row or column is left uncrossed out, stop .otherwise, move to the cell to the right if acolumn has just been crossed out or below if a row has been crossed out .Go to step (i). [6]
5.2. Minimum-Cost Method
The minimum-cost method finds a better starting solution by concentrating on the cheapest routes. Themethod starts by assigning as much as possible to the cell with the smallest unit cost. Next, the satisfiedrow or column is crossed out and the amounts of supply and demand are adjusted accordingly. If both arow and a column are satisfied simultaneously, only one is crossed out, the same as in the northwest corner method .Next ,look for the uncrossed-out cell with the smallest unit cost and repeat the processuntil exactly one row or column is left uncrossed out . [6]
5.3 Vogels Approximation Method (VAM)
Vogels Approximation Method is an improved version of the minimum-cost method that generallyproduces better starting solutions.(i) For each row (column) determine a penalty measure by subtracting the smallest unit cost element inthe row (column) from the next smallest unit cost element in the same row (column).(ii) Identify the row or column with the largest penalty. Break ties arbitrarily. Allocate as much aspossible to the variable with the least unit cost in the selected row or column .Adjust the supply anddemand and cross out the satisfied row or column. If a row and column are satisfied simultaneously,only one of the two is crossed out, and the remaining row (column) is assigned zero supply (demand).(iii) (a) If exactly one row or column with zero supply or demand remains uncrossed out, stop.(b) If one row (column) with positive supply (demand) remains uncrossed out, determine the basicvariables in the row (column) by the least cost method .stop.(c) If all the uncrossed out rows and columns have (remaining) zero supply and demand, determine thezero basic variables by the least-cost method .stop. ).[6]
(d) Otherwise, go to step (i).
5.4 Row Minimum Method
Row minimum method start with first row and choose the lowest cost cell of first row so that either thecapacity of the first supply is exhausted or the demand at jth distribution center is satisfied or both. Threecases arise:
March Issue Page 94 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
7/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
(i)If the capacity of the first supply is completely exhausted, cross off the first row and proceed to thesecond row.(ii)If the demand at jth distribution center is satisfied, cross off the j th column and reconsider the first rowwith the remaining capacity.(iii)If the capacities of the first supply as well as the demand at j th distribution center are completelysatisfied, make a zero allocation in the second lowest cost cell of the first row .cross off the row as wellas the jth column and move down to the second row.Continue the process for the resulting reduced transportation table until all the rim conditions (supplyand demand condition) are satisfied. [8]
5.5 Column Minimum Method
Column minimum method starts with first column and chooses the lowest cost cell of first column sothat either the demand of the first distribution center is satisfied or the capacity of the i th supply isexhausted or both .three cases arise:(i)If the demand of the first distribution center is satisfied, cross of the first column and move right to thesecond column.(ii)If the capacity of ith supply is satisfied, cross off ith row and reconsider the first column with theremaining demand.(iii)If the demands of the first distribution center as well as the capacity of the i th supply are completely
satisfied, make a zero allocation in the second lowest cost cell of the first column. Cross of the columnas well as the ith row and move right to the second column.Continue the process for the resulting reduced transportation table until all the rim conditions aresatisfied.
6. ILLUSTRATIVE EXAMPLE (SUNRAY TRANSPORTATION)
Sun Ray Transportation Company ships truckloads of grain from three silos to four mills [6]. Thesupply (in truckloads) and the demand (also in truckloads) together with the unit transportation costs pertruckload on the different routes are summarized in the transportation model in table 1.
Table 1. Transportation model of example (SunRay Transportation)
The model seeks the minimum-cost shipping schedule between the silos and the mills. This is equivalentto determining the quantity xij shipped from silo i to mill j (i=1, 2, 3; j=1, 2, 3, 4)
6.1. Northwest-Corner method
The application of the procedure to the model of the example gives the starting basic solution in table.2.
March Issue Page 95 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
8/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
Table 2. The starting solution using Northwest-corner method
The Starting basic Solution is given asx11=5, x12=10, x22=5, x23=15, x24=5, x34=10The objective function value isZ=510+102+57+159+520+1018=$520
6.2 Minimum Cost Method
The minimum-cost method is applied to Example (SunRay Transportation) in the following manner:1. Cell (1,2) has the least unit cost in the tableau (=$2).the most that can be shipped through (1,2) isx12=min (15,15)=15 which happens to satisfy both row 1 and column 2 simultaneously, we arbitrarilycross out column 2 and adjust the supply in row 1 to 0.2. Cell (3,1) has the smallest uncrossed-out unit cost (=$4).Assign x 31=5, cross out column 1 because itis satisfied and adjust the demand of row 3 to 10-5=5 truckloads.3.continuing in the same manner ,we successively assign 15 truckloads to cell (2,3), 0 truckloads to cell(1,4), 5 truckloads to cell(3,4) and 10 truckloads to cell (2,4).The resulting starting solution is summarized in this table.3. The arrows show the order in which theallocations are made.
The starting solution (consisting of 6 basic variables) isx12=15, x14=0, x23=15, x24=10, x31=5, x34=5The associated objective function value isZ=152+011+159+1020+5+518 = $475The quality of the least cost starting solution is better than of the northwest corner method because ityields a smaller value of Z ($475 versus $520 in the north-west corner method.
Table 3. The starting solution using minimum-cost method
March Issue Page 96 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
9/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
6.3 Vogels Approximation Method (VAM)
VAM is applied to Example (SunRay Transportation) in the following manner
Table 4. Step1 to determine the starting solution using (VAM)
1. We computes from row 3 because row 3 has the largest penalty =10 and cell (3,1) has the smallestunit cost in the row, the amount 5 is assigned to x31.Column 1 is now satisfied and must be crossed out.Next, new penalties are recomputed in the table. 5.2. Shows that row 1 has the highest penalty(=9).hence ,we assign the maximum amount possible to cell(1,2),which yields x12=15and simultaneously satisfies both row 1 and column 2 .we arbitrarily cross outcolumn 2 and adjust the supply in row 1 to zero as table. 6.
Table 5. Step2 to determine the starting solution using (VAM)
Table 6. Step3 to determine the starting solution using (VAM)
March Issue Page 97 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
10/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
3.Continuing in the same manner, row 2 will produce the highest penalty (=11) and we assign x23=15,which crosses out column 3 and leaves 10 units in row 2 .Only column 4 is left ,and it has a positivesupply of 15 units . Applying the least-cost method to that column, we successively assign x 14=0,x34=5and x24=10 (verify!) as table. 7.
Table 7. Step 4 to determine the starting solution using (VAM)
The associated objective function value isZ=152 + 011 + 159 + 1020 + 54 + 518=$475.VAM produces a better starting Solution.
6.4 Row Minimum Method
Row minimum method is applied to example (SunRay Transportation) in table. 8.
Table 8. The starting solution using row minimum method
March Issue Page 98 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
11/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
The associated objective function value isZ = 152+125+915+205+1018 = $505This cost is less than northwest-corner method
6.5 Column Minimum Method
Column minimum method is applied to Example (SunRay Transportation) in table. 9.The associated objective function value isZ =45 +152 +915 +1020 +185 =$ 475.This cost is less than northwest-corner method.
Table 9. The starting solution using column minimum method
6.6 COMPARISON BETWEEN THE FIVE METHODS
North-west corner method is used when the purpose of completing demand No. 1 and then the nextand is used when the purpose of completing the warehouse No. 1 and then the next. Advantage ofnorthwest corner method is quick solution because computations take short time but yields a badsolution because it is very far from optimal solution. Vogel's approximation method and Minimum-costmethod is used to obtain the shortest road. Advantage of Vogels approximation method and Minimum-
cost method yields the best starting basic solution because gives initial solution very near to optimalsolution but the solution of Vogels approximation methods is slow because computations take longtime.
The cost of transportation with Vogel's approximation method and Minimum-cost method is less thannorth-west corner method. Row-minimum method is used when the purpose of completing thewarehouse No. 1 and then the next. Row minimum cost is useful in small number of supply and whenthe cost of transportation on supply. The cost of transportation is less than North-west corner method.Column minimum method is used when the purpose of completing demand no.1 and then the next.Column minimum cost is useful in small number of demand and when the cost of transportation ondemand. The cost of transportation is less than North-west corner method.
7. Transportation Network Data
We present experiments with six months of origin destination (OD) data from a large third-partylogistic company. The dataset consists of 98,292 transactions. Each transaction has 11 attributes,described in Table 10. There are 4038 distinct latitude-longitude (LL) pairs in the dataset, with 1797distinct origins and 3770 distinct destinations (several locations are both). The dataset contains 20, 900distinct OD pairs (i.e., there are often multiple deliveries between the same origin and destination overthe six months). The edges are labeled with the other attributes of the transaction: pickup date, deliverydate, distance, hours, weight, and mode.
March Issue Page 99 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
12/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
Labeling edges with the exact values would lead to few frequent patterns being detected, since theedge labels are often unique. Instead, we use a binning strategy. Each label (distance, hours, weight) isdivided into ranges, giving a few distinct labels for each type (we used seven for gross weight and tenfor transit hours in the experiments reported in this paper). As a result, even edges with similar (thoughnot exactly equal) distances, times, and weight are considered to support the same pattern. Since therange of values is quite large, this makes perfect sense. For example, the range for weight is about 500tons - in this case, two different transactions from the same source to the same destination with weights49 tons and 52 tons respectively should be considered equal. Binning facilitates this. This dataset isnaturally represented as a directed graph by mapping locations to vertices. Each transaction can then berepresented as the edge of an OD pair. The dataset does give a fully connected graph. Minimum,maximum, and average out-degrees are 1, 2373, and 12 respectively, and Minimum, maximum, andaverage in-degrees are 1, 832, and 6.
We generate three different graphs from the data, named OD GW,OD TH,OD TD. Each graph has thesame set of vertices and edges but different labeling scheme for the edges. OD GW uses GROSSWEIGHT , OD TH uses MOVE TRANSIT HOURS, while OD TD uses TOTAL DISTANCE for edgelabeling.
Table 10. Transportation Network Data Description
Name DescriptionID Unique transaction identifier.REQ PICKUP DT Requested date to pick up the load.REQ DELIVERY DT Requested delivery date.ORIGIN LATITUDE Latitude of source (to nearest 0.1 degree.)ORIGIN LONGITUDE Longitude of source (to nearest 0.1 degree.)DEST LATITUDE Latitude of destination (to nearest 0.1 degree.)DEST LONGITUDE Longitude of destination (to nearest 0.1 degree.)TOTAL DISTANCE Road miles between origin and destination.GROSS WEIGHT Weight of load.MOVE TRANSIT HOURS Hours needed to get from origin to destination.TRANS MODE Truckload or Less than Truckload.
In the following two sections, we define two valuable patterns (with distinct underline meanings) fordecision making in transportation and logistic domains. We propose first cut approach that utilizes theexisting graph mining software to discover these patterns. From experimental results, although theexisting software are not suitable to mine these patterns and the first cut approach is limited, it doespresent the need for further research in this area.
7.1 Structurally Similar Routes
One problem we have addressed is identifying self similarity within the transportation network. The goalis to identify structurally similar patterns that occur in many locations. For example, a pattern might be abow-tie with several small loads converging on a location, large loads to a distant location, and smallloads converging on the distant location. Seeing this pattern, a transportation expert could find a way tobetter utilize resources outside the bounds of traditional optimization methods. Instead of just optimizing
truck routes, the company could use multi-modal transportation, placing trailers on rail cars for the largeload long distance portion of the pattern, and using the rail-capable trailers as a pool for the shortdeliveries in the vicinity of the endpoints. This is just a hypothetical example; best utilizing thediscovered patterns requires considerable external knowledge. The key is that since the patterns arefrequent, innovative transportation approaches that optimize deliveries in those patterns can be appliedin many places. Since we are interested only in structural similarity and not particular locations, weassign all vertices the same label. Thus, vertex labeling is a non-factor in finding frequent sub-graphs.The three variants for edge labels; weight, distance, and time; were described earlier.
March Issue Page 100 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
13/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
7.2 Classification based experiments
We first ran Wekas version of the C4.5 algorithm (J4.8) on the discretized dataset. The J4.8 modelwas 96% accurate in classifying instances based on the class attribute TRANS MODE {LTL, TL}. Theclassification tree first splits on the GROSS WEIGHT attribute, which is consistent with the strongassociation rules generated by the Apriori algorithm in section 7.1. J4.8 was also run on the discretized
dataset with the TRANS MODE attribute removed and TOTAL DISTANCE set as the class attribute.The results from this experiment were interesting in the sense that TOTAL DISTANCE and MOVETRANSIT HOURS were not as highly correlated as either TOTAL DISTANCE and DESTINATIONLATITUDE or TOTAL DISTANCE and ORIGIN LATITUDE.
7.3 Clustering based experiments
We used the original undiscretized data set as the training set input to the EM (expectation-maximization) algorithm. The algorithm works by assigning each object to a cluster based on a weightrepresenting the probability of membership. Figure 5 summarizes the results produced by the EMalgorithm. The training data were split into nine clusters varying in size from 3 instances in cluster 0 to19,386 instances in cluster 2.
Figure 5. Clustering statistics for transportation data
Figure 6(a) reflects the mean TOTAL DISTANCE within each cluster. Similarly Figure 6(b) reflects themean TRANSIT HOURS within each cluster. As is evident in the figures, Cluster 0 contains the outliersin this data set. These three shipments have on average, traveled over3,000 miles in less than 24 hours.Looking at the latitude and longitude of the origin and destination points (as well as the gross weight),one can conclude the shipments are handled as air freight (originating in the Pacific Northwest anddelivered to Hawaii).
0
500
1000
1500
2000
2500
3000
3500
0 1 2 3 4 5 6 7 8
cluster number
total_
distance
Figure 6(a). Cluster Comparison(total_distance).
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8
cluster number
tra
nsit_
hours
Figure 6(b). Cluster comparison(transit_hours)
March Issue Page 101 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
14/19
7
Similar analyses can be applied to characterize the remaining clusters. Clusters 2, 5, 7 and 8 can begrouped together and labeled short-haul. Similarly, clusters 1, 3, 4, and 6 can be referred to as long-haul routes. Traditional data mining techniques have produced interesting and meaningful results tosummarize our data despite the exclusion of two critical temporal attributes. Further experimentationwith traditional mining algorithms is required to explore the potential and limitations of these techniqueson temporal transportation network data. In addition, from these data mining results, it is obvious thatthe insights from the structural characteristics of the data cannot be derived. As a consequence,conventional methods can only produce limited insights on this transportation and logistic dataset.
8. Logistics performance at companies
The productivity indexes of logistics are quantitative measures. Their amounts just partly guarantee thehigh quality but e.g. the transport speed is an important factor of transport quality. For this reason, itsperformance matrix is worth being looked at.
Table 11. Performance matrix
Names Managing
(Dispatching)
Storing
Packaging
Material handling
Transport
(In & Out)
1.Personnel Commitment(staff)
Vol. loaded in/out(Staff)
Packaged pcs(Staff)Handled volume.(Staff)
Tr.edvolume(Staff)
2.CompanyFacilities
Fac.work time(fac.time capacity)
Temporaryutilizationof facilities (%)
Vol. loaded in/outfac.and year
Packaged pcs (facilitiesand year)
Handled vol(fac.and year)
Ton(km)vehicle and year
Transportedvol.(vehicle andyear)Run km(Vehicleand Year)Run km(Vehicleand Year)
Averagetransportationdistance(km)Utilizationof vehicles(%)related to time,load capacity andtkm
3.Space
and surface
Utilization of loading
surface (%)Utilization of loadingspace (%)
March Issue Page 102 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
15/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
4.KeepingInventory
Av. Inventory(volume, value)Av. storage time(day, hour, minute)Av. inventoryefficiency (day,minute)Rotationfrequency/yearCommitments/yearAv. time of comm.(hour, minute)
5.Average andtotal
Numberof comm. /yearNumbers ofsuppliers/yearNo of buyers/yearOp. time/comm.Return/year
Handled. vol (staff)
Handled vol.(Year)
Packaged pcs(staff)
Packaged pcs(year)
Transported vol.(Year)
No of transports(Year)
Av vol.(transport)Av.time(transport)
8.1Costs of logistics
Along the logistics chain there are seller-buyer relations in every joint point. The seller adds someprofit to his costs and offers his service at a price to the buyer. For the buyer the previous price meanscosts again, he adds more value and this way some profit to the previous service and sells it. Thisrelation is repeated in the whole logistics process till the final buyer (customer).It is not easy to separatelogistics costs from the whole production costs for a product which has many phases.
For this reason we should take a simple industrial company which gets the raw material immediatelyfrom the extractive industry, manufactures just one kind of products. This product will be sold to thecustomers (not to another industry). Insurance is always included .This scheme is repeated on the wholelogistics chain.
Table 12. Contracted logistics costs matrix (according to the kinds of costs)
Names Managing
(Dispatching)
Store aging
Packaging
Mat. handling
(own &foreign)
Transport
(own & external)
Costs accordingto cost places
Share of total
logistics costs
Managing costsCommitment
Total man. costs
total log. costs
Stor.,pack,handl.costsvolume,unit,value
Total stor & inv.coststotal log. costs
Tr.costscommitmentTr.costsvolume,tkm,km
Tr. coststime& vehicleTotal tr.costsTotal log. Costs
Share of costgroups in totallogistics costs
Personnel costs/total log. CostsCompany log. fac. costs/total log. costsSpace and surface costs/total log. costsInventory costs/total log. costs
March Issue Page 103 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
16/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
Foreign log. costs/total log. Costs
Costs of totallogisticsperformances
Total log. costs/yearTotal log. costs/productTotal log. costs/total production costs
We have to remember the quality costs of logistics. There are here three groups of quality costs:
- failure costs (wastes),- control and assessment costs,- prevention costs.High quality performance does not necessarily mean higher costs. The costs of wastes can be rearranged,reduced or eliminated by quality control and prevention.How to save logistics expenditures at macroeconomic level:- reducing the number of logistics activities- optimum selection of logistics technologies, facilities (packaging, load unit, multimodal transport,preferring rail- and waterways etc.)- restructuring industry allocation- modern, comprehensive organisation (logistics centers, just in time etc.)With the specialisation and globalization of production and with ensuring a wide range of choice ofconsumer's goods, the role of logistics is getting more and more emphasized. The rationalisation oflogistics process is essential.
9. RESULTS
OBJECT-ORIENTED PROGRAMMING
Object-Oriented Programming (OOP) is method of implementation in which programs are organized ascooperative collections of objects and each object is an instance of some classes, classes are related toone another via inheritance relationship. [5] In Object-Oriented programming, the data and functions areintegrated. An object is like a box containing it data and its functions which can operate on the data.[4]Object-Oriented programming languages Provides great flexibility, clarity and reusability throughinheritance. It leads to faster software development, increased quality, easier maintenance, and flexiblemodifiability. Objects are the basic elements for executing object oriented programs while classes arethe basic elements for defining object-oriented programs. If any of these elements is missing, it is not an
object-oriented program.[4] object-oriented programming languages such as java, C# and C++.
SOLVING TRANSPORTATION PROBLEM USING JAVA
We need to describe the five methods (mentioned above) of transportation model in LP using the fivealgorithms and we draw a flow chart for each algorithm.After designing algorithms for the five methods we develop JAVA program for each one. We used Javalanguage to facilitate getting the result and the complex problems which take long time using LPsolution. After running these programs we compared between each solution using Java program and LPsolution which show that have the same result and compare between different solutions for choosing lessvalue of the objective function.The main ideas from design five Java programs are save time, money, and effort In the example (Sun Ray Transportation) we use the five Java programs to minimize the cost of
transportation and determine the number of units transported from source i to destination j.The results are shown as follows. The result of northwest-corner method program by Java, the cost of transportation =$520The number of units transported from source i to destination j.we transportsupply [0] to demand[0] =5supply [0] to demand[1] =10supply [1] to demand[1] =5
March Issue Page 104 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
17/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
supply [1] to demand[2] =15supply [1] to demand[3] =5supply [2] to demand[3] =10Press any key to continue The result of minimum-cost method program by Java, the cost of transportation =$475The number of units transported from source i to destination jwe transportsupply [0] to demand[1] =15supply [1] to demand[2] =15supply [1] to demand[3] =10supply [2] to demand[0] =5supply [2] to demand[3] =5Press any key to continue The result ofVogels approximation method program by Java language, the cost of transportation= $475
The number of units transported from source i to destination jwe transportsupply [0] to demand[1] =15supply [1] to demand[2] =15supply [1] to demand[3] =10
supply [2] to demand[0] =5supply [2] to demand[3] =5Press any key to continue The result of row minimum method program by Java language, the cost of transportation =$505The number of units transported from source i to destination jwe transporttransport supply [0] to demand[1] =15transport supply [1] to demand[0] =5transport supply [1] to demand[2] =15transport supply [1] to demand[3] =5transport supply [2] to demand[3] =10Press any key to continue The result of column minimum method program by Java language, the cost of transportation =$475
The number of units transported from source i to destination jwe transportsupply [0] to demand[1] =15supply [1] to demand[2] =15supply [1] to demand[3] =10supply [2] to demand[0] =5supply [2] to demand[3] =5Press any key to continueMinimum-cost method, Vogels approximation method and column minimum methods are having thesame objective value are equal to $ 475 and give less value from other methods. We choose less resultfrom these results to reduce the cost of transportation andwe transportsupply [0] to demand[1] =15
supply [1] to demand[2] =15supply [1] to demand[3] =10supply [2] to demand[0] =5supply [2] to demand[3] =5The results of the five programs using java are equal to LP solution but the solution using Java languagefaster and easier than LP solution.
March Issue Page 105 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
18/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
10. CONCLUSION
Running the five Java programs for solving transportation problem shows that the result of the fiveJava programs is equal to the result of the LP solution. But the result of the five programs are different.The decision maker may choose the optimal result by the running of the five programs (minimum) anddetermine the number of units transported from source i to destination j
Logistics costs take 18-23% of total production costs (in a wider sense even 40%).For this reasonproduction companies are going to make these activities more effective. Big companies have thirdpartners (forwarders) to make it. A forwarder can comprehend and optimize a longer interval of logisticschain.
The share of distribution and production logistics is not right. The border is not sharp. In the case ofquality the question is how and not where. The seller and buyer relation can be found in all joint pointsof the logistics chain.
REFERENCES
[1] K. Bansal, S. Vadhavkar, and A. Gupta, Neural networks based data mining applications formedical inventory problems, International Journal of Agile Manufacturing, 1(2):187200, 1998.[2] K. Bansal, S. Vadhavkar, and A. Gupta, Neural networks based forecasting techniques for inventory
control applications, Data Mining and Knowledge Discovery, 2(1):97102,1998.[3] Liu Xu, Guojun Mao. An Algorithm to Approximately Mine Frequent Closed Itemsets from DataStreams, Acta Electronica Sinica. 2007.5[4] Peretz Shovel, Functional and object-oriented analysis and design (an integrated methodology),Idea Group Publishing (an imprint of Idea Group Inc.), United States of America, 2006.[5] Grady Booch, Object-oriented Analysis and design, Addison-Wesley Professional, 2 editions, USA,1993.[6] Hamdy A.Taha, Operations Research: An Introduction, Prentice Hall, 7 editions 5 ,USA,2006.[7] Prem Kumar Gupta, D.S.Hira, Operations Research: An Introduction, S.Chand and Co., Ltd. NewDelhi, 1999.[8] Reghu Ramakrishnan, Johannes Gehrke, Database management systems, second Edition, McGraw-Hill, August 1999.[9] The IEA and transport, Feb. 27 2003.
March Issue Page 106 of 107 ISSN 2229 5216
7/29/2019 Data Mining Analytics to Minimize Logistics Cost
19/19
International Journal of Advances in Science and Technology,Vol. 2, No.3, 2011
Abraham Paul did his B.Sc. (Computer Science) from the University of Madras,M.C.A at Karunya Institute of Technology (now Karunya University), Coimbatore.He is working with VIT University for the past 10 years. He has one internationaljournal publication and two papers in inter-national conference. His areas of interestare Software Engineering and Data Mining.
Dr. V Saravanan obtained his Bachelors degree in Mathematics from University ofMadras during 1996 and Masters Degree in Computer Applications from Bharathiar University during1999. He has completed his PhD in Computer Science in the Department of Computer Science andEngineering, Bharathiar University during 2004. He specialized on automated and unified data miningusing intelligent agents. His research area includes data warehousing and mining, software agents and
cognitive systems. He has presented many research papers in National, International conferences andJournals and is also guiding many researchers leading to their PhD degree. He has totally 10 yearsexperience in teaching including 3 years as researcher in Bharathiar University. He is the life member ofComputer Society of India, Indian Society for Technical Education, and Indian Association of Researchin Computing Sciences and and International Association of Computer Science and InformationTechnology. He worked as Professor & HOD of the Department of Computer Applications in KarunyaUniversity, Coimbatore from 1999-2009. At present, Professor & Director, Department of ComputerApplications, Dr. NGP Instituite of Technology, Coimbatore, Tamilnadu.
Dr. P. Ranjit Jeba Thangaiah did his B.Sc (Physics) at P.S.G. College of Arts andScience, Coimbatore, M.C.A at Karunya Institute of Technology (now Karunya University),Coimbatore, M.Phil. at Manaonmanian Sundaranar University, Tirunelveli, He did his full time Ph.D. atBharathiar University. He is working with Karunya University for the past 9 Years. He has published 5papers in International Journals and 3 papers in International Conferences. His areas of Interest are datamining and machine learning.