+ All Categories
Home > Documents > Soft Computing For Controle

Soft Computing For Controle

Date post: 23-Feb-2016
Category:
Upload: gizela
View: 28 times
Download: 0 times
Share this document with a friend
Description:
Evolving Fuzzy Rules with Genetic Programming and Clustering. Soft Computing For Controle. G-REX (Previous work). The transformation of an highly accurate opaque model to a comprehensible model . Genetic programming Black box Arbitary representation and fitness function - PowerPoint PPT Presentation
Popular Tags:
25
Soft Computing For Controle Evolving Fuzzy Rules with Genetic Programming and Clustering
Transcript
Page 1: Soft Computing For Controle

Soft Computing For Controle

Evolving Fuzzy Rules with Genetic Programming and Clustering

Page 2: Soft Computing For Controle

The transformation of an highly accurate opaque model to a comprehensible model.

Genetic programming Black box Arbitary representation and fitness function

Balances Accuracy and Comprehensibility

G-REX (Previous work)

Accept Reject

IF Salary > 5000

IF Age > 25

Reject

X1 X2 Y

1 4 1

4 3 1

2 1 14 5 0

5 2 0

X1 X2 Z

1 4 1

4 3 1

2 1 04 5 0

5 2 0

Page 3: Soft Computing For Controle

GP - Crossover

Page 4: Soft Computing For Controle

Background

Evolving Fuzzy Decision Trees With Genetic Programming and Clustering

J. Eggermont, (2001) Automatic fuzzyfication using K-Means Genetic Programming Fuzzy Representation

Page 5: Soft Computing For Controle

Membership functions

Three types of membership function Distances does not need to be equal Based on medioids/centroids

Page 6: Soft Computing For Controle

Membership functions

Page 7: Soft Computing For Controle

K-means

Most frequently used clustering method Fast, deterministic and easy to implement.

J.B MacWueen (1967) K- stand for the number of clusters

Each cluster is represented by one membership function A cluster is represented by a centroid.

The mean value of the members An instance belongs to the closest centroid

Page 8: Soft Computing For Controle

1 Euclidian distance

Page 9: Soft Computing For Controle

2 The new centroid is

the mean of its members

Page 10: Soft Computing For Controle

3

Recalculate members Repeat until no change

Page 11: Soft Computing For Controle

Kaufmans Initialization

Step 1. The instance closest to the mean valueStep 2-3 Choose a instance far away from the other medioids with many

instance close by.

K-Means is sensitive to the initialization method Pêna J.M. Lozano J. A. and Larranga P. (1999)

An Empirical Investigation of Four Initialization Methods for the K-Means Algorithm

Page 12: Soft Computing For Controle

Three types of membership function Distance does not need to be equal Based on medioids

Membership functions

Page 13: Soft Computing For Controle

GP Representation

All variables with less than k unique values are treated as crisp sets.

Page 14: Soft Computing For Controle

Representation

Page 15: Soft Computing For Controle

Calculating membership values

Page 16: Soft Computing For Controle

Fitness function

2

1 1

11 ( )r n

Brier ij ijj i

fitness f En

Not precise enough

Reward is equal to the membership Value for the correctly predicted instance

1- the MSE of each membership function

Page 17: Soft Computing For Controle

Experiments

5 classification datasets Only continuous variables

IRIS, WINE Categorical and continuous

COLIC, CLEAVLAND, PIMA 10-fold cross validation Stratification Fuzzy GP vs standard GP (if rules) Evaluated against

Accuracy (ACC) Area under ROC-curve (AUC) Brier Score (BRI)

Page 18: Soft Computing For Controle

Results

DatasetFuzzy

ACC AUC BRI  Train Test Train Test Train Test Size

 IRIS 96.8 96.0 99.5 99.1 7.8 8.7 7.4 CLEAVLAND 76.2 75.8 81.3 82.6 37.7 36.7 8.6 WINE 89.4 90.4 97.6 98.1 15.3 13.9 9.0 COLIC 67.2 66.3 66.1 64.6 48.4 50.5 9.0 PIMA 89.4 90.4 97.6 98.1 15.3 13.9 9.0

DatasetIF

ACC AUC BRI  Train Test Train Test Train Test Size

 IRIS 96.2 93.3 98.1 96.5 7.0 12.2 11 CLEAVLAND 76.8 72.5 76.6 72.4 35.6 40.3 6 WINE 90.8 87.5 95.6 91.2 16.3 22.6 11 COLIC 81.5 81.5 81.9 77.0 28.9 29.2 6 PIMA 90.8 87.5 95.6 91.2 16.3 22.6 11

Page 19: Soft Computing For Controle

Iris

Page 20: Soft Computing For Controle

Wine

Page 21: Soft Computing For Controle

Horse Colic

Page 22: Soft Computing For Controle

PIMA Diabetes

Page 23: Soft Computing For Controle

Cleveland (Heart disease)

Page 24: Soft Computing For Controle

Disscussion

Current membership function removes information from the variable A way to handle outliers

Some extremely simply if rules are better for some dataset. Categorical variables Should not be used as only method

Easy to remember rules but how accurate will they be as a decision support?

Gives a comprehensible explanation that could ad trust and there by improve predictions.

Page 25: Soft Computing For Controle

Future work

Alternative membership function

Fuzzy regression

?


Recommended