On the Combination of Accuracy and Diversity Measures for Genetic Selection of Bagging Fuzzy Rule-Based Multiclassification Systems
Krzysztof Trawiński, Arnaud Quirin, Óscar Cordón · 2009
A preliminary study combining two diversity measures with an accuracy measure in two bicriteria fitness functions to genetically select fuzzy rule-based multiclassification systems is conducted in this paper. The fuzzy rule-based classification system ensembles are generated by means of bagging and mutual information-based feature selection. Several experiments were developed using four popular UCI datasets with different dimensionality in order to analyze the accuracy-complexity trade-off obtained by a genetic algorithm considering the two fitness functions. Comparison are made with the initial fuzzy ensemble and a single fuzzy classifier.