Selecting a Small Subset of Informative Genes from Gene Expression Data by Using a Modified Binary Particle Swarm Optimisation

Gene expression technology, especially microarrays, can be used to measure the expression levels of thousands of genes simultaneously in biological organisms. Gene expression data produced by microarrays are expected to be useful for cancer classification. To select a small subset of informative genes for cancer classification, many researchers have analysed the gene expression data using various computational intelligence methods. However, due to the small number of samples compared with the huge number of genes (high-dimensional data), irrelevant genes, and noisy genes, many of the computational methods face difficulties in selecting the small subset. Thus, we propose a modified binary particle swarm optimisation to select a small subset of informative genes that are relevant for the cancer classification. In the proposed method, we introduce the particle speed and a rule for increasing the probability of bits in a particle’s position to be zero. The method was empirically applied to a suite of four well-known benchmark gene expression data sets. The experimental results demonstrate that the proposed method outperforms the conventional version of binary particle swarm optimisation (BPSO) and other related works in terms of classification accuracy and the number of selected genes. In addition, this method also produces lower running times compared to BPSO.