Modular feature selection using relative importance factors

Abstract

Feature selection plays an important role in finding relevant or irrelevant features in classification. Genetic algorithms (GAs) have been used as conventional methods for classifiers to adaptively evolve solutions for classification problems. In this paper, we explore the use of feature selection in modular GA-based classification. We propose a new feature selection technique, Relative Importance Factor (RIF), to find irrelevant features in the feature space of each module. By removing these features, we aim to improve classification accuracy and reduce the dimensionality of classification problems. Benchmark classification data sets are used to evaluate the proposed approaches. The experiment results show that RIF can be used to determine irrelevant features and help achieve higher classification accuracy with the feature space dimension reduced. The complexity of the resulting rule sets is also reduced which means the modular classifiers with irrelevant features removed will be able to classify data with a higher throughput.