Hybrid Search Optimization Techniques for the Classification of Cancer Data from the Optimal Set of Features

Abstract

The paper seeks to find cancer-related data, with the help of the dataset's reduced features. The dataset is searched for globally optimal features using particle swarm optimization. It typically produces higher classification results, but on occasion, these algorithms choose neighborhood features incorrectly, creating a local optimal feature selection trap. In order to select the best features from the microarray gene expression data, this work suggests a method that combines the two algorithms with fuzzy rough set to form Particle Swarm Optimization and Tabu Search with Fuzzy Rough Set for Optimal feature selection (PSTFRO) technique. It simultaneously selects neighborhood features and global features. The usefulness of the suggested selection of features method is evaluated in relation to receiver operating characteristics, exact classification, computation speed, and a positive predicted value using the fuzzy rough nearest neighbor classifier. The proposed approach produces better results than the presently used global optimal, local optimal, and optimal feature selection algorithms.