A Constraint and Rule in an Enhancement of Binary Particle Swarm Optimization to Select Informative Genes for Cancer Classification

Abstract

Gene expression data have been analyzing by many researchers by using a range of computational intelligence methods. From the gene expression data, selecting a small subset of informative genes can do cancer classification. Nevertheless, many of the computational methods face difficulties in selecting small subset since the small number of samples needs to be compared to the huge number of genes (high-dimension), irrelevant genes and noisy genes. Hence, to choose the small subset of informative genes that is significant for the cancer classification, an enhanced binary particle swarm optimization is proposed. Here, the constraint of the elements of particle velocity vectors is introduced and a rule for updating particle's position is proposed. Experiments were performed on five different gene expression data. As a result, in terms of classification accuracy and the number of selected genes, the performance of the introduced method is superior compared to the conventional version of binary particle swarm optimization (BPSO). The other significant finding is lower running times compared to BPSO for this proposed method.