Designing Fuzzy Ensemble Classifiers by Evolutionary Multiobjective Optimization with an Entropy-Based Diversity Criterion

Authors:
Yusuke Nojima;Hisao Ishibuchi
Affiliations:
Osaka Prefecture University, Japan;Osaka Prefecture University, Japan
Venue:
HIS '06 Proceedings of the Sixth International Conference on Hybrid Intelligent Systems
Year:
2006

Citing 0
Cited 5

Genetic rule selection with a multi-classifier coding scheme for ensemble classifier design

International Journal of Hybrid Intelligent Systems - Hybridization of Intelligent Systems
Effects of Diversity Measures on the Design of Ensemble Classifiers by Multiobjective Genetic Fuzzy Rule Selection with a Multi-classifier Coding Scheme

HAIS '08 Proceedings of the 3rd international workshop on Hybrid Artificial Intelligence Systems
Review Article: Multi-objective nature-inspired clustering and classification techniques for image segmentation

Applied Soft Computing
A novel framework to design fuzzy rule-based ensembles using diversity induction and evolutionary algorithms-based classifier selection and fusion

IWANN'13 Proceedings of the 12th international conference on Artificial Neural Networks: advances in computational intelligence - Volume Part I
Multiobjective genetic classifier selection for random oracles fuzzy rule-based classifier ensembles: How beneficial is the additional diversity?

Knowledge-Based Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we propose a multi-classifier coding scheme and an entropy-based diversity criterion in evolutionary multiobjective optimization algorithms for the design of fuzzy ensemble classifiers. In a multi-classifier coding scheme, an ensemble classifier is coded as an integer string. Each string is evaluated by using its accuracy and diversity. We use two accuracy criteria. One is the overall classification rate of the string as an ensemble classifier. The other is the average classification rate of component classifiers in the ensemble classifier. As a diversity criterion, we use the entropy of outputs from component classifiers in the ensemble classifier. We examine four formulations based on the above criteria through computational experiments on benchmark data sets in the UCI machine learning repository. The experimental results show the effectiveness of the multi-classifier coding scheme and the entropy-based diversity criterion.