Effects of Data Reduction on the Generalization Ability of Parallel Distributed Genetic Fuzzy Rule Selection

Genetic fuzzy rule selection has been successfully used to design accurate and interpretable fuzzy classifiers from numerical data. In our former study, we proposed its parallel distributed implementation which can drastically decrease the computational time by dividing both a population and a training data set into sub-groups. In this paper, we examine the effect of data reduction on the generalization ability of fuzzy rule-based classifiers designed by our parallel distributed approach. Through computational experiments, we show that data reduction can be realized without severe deterioration in the generalization ability of the designed fuzzy classifiers.