Sampling strategies and sample redundancy in population-based fuzzy modeling and classification

Training set sampling strategies are used in the context of population-based fuzzy modeling and classification to evaluate the quality of a data set. The objective is to explain the use of down-sampling strategies as a means for reducing the number of redundant samples. It is believed the clustering to be rather generally applicable to sampling in classification applications. The hypothesis is empirically validated by examining the performance of differential evolution fuzzy classifiers on Iris and Breast Cancer Wisconsin data sets. The learning curves of the classifiers are analyzed with respect to the choice of the sampling strategies. A nearly-optimal classification performance of the classifier is achieved using a relatively small training sample, showing that population-based fuzzy modeling and classification can be successfully applied to large data sets with affordable computation.