Cancer classification using a novel gene selection approach by meansof shuffling based on data clustering with optimization

Abstract

This research presents an innovative method for cancer identification and type classification usingmicroarray data. The method is based on gene selection with shuffling in association with optimiza-tion based unconventional data clustering. A new hybrid optimization algorithm, COA-GA, is developedby synergizing recently invented Cuckoo Optimization Algorithm (COA) with a more traditional geneticalgorithm (GA) for data clustering to select the most dominant genes using shuffling. For gene classi-fication, Support Vector Machine (SVM) and Multilayer Perceptron (MLP) artificial neural networks areused. Literature suggests that data clustering using traditional approaches such as K-means, C-means andHierarchical do not have any impact on classification accuracy. This is also confirmed in this investiga-tion. However, results show that optimization based clustering with shuffling increase the classificationaccuracy significantly. The proposed algorithm (COA-GA) not only outperforms COA, GA and ParticleSwarm optimization (PSO) in achieving better classification performance but also reaches a better globalminimum with only few iterations. Higher accuracy is observed to have achieved with SVM classifiercompared to MLP in all datasets used.