Gene Selection for Diagnosis of Cancer in Microarray Data Using Memetic Algorithm

摘要

Selecting a small subset of genes that helps to build a good classification model for prediction of disease on the microarray data is a very demanding optimization problem. Genetic algorithm (GA) is a population-based optimization algorithm, which has a lot of applications in the field of molecular biology. But the premature convergence is one of the limitations of GA. Memetic algorithm (MA), an extension of GA, diminishes the possibility of such premature convergence. Microarray technology enables to measure the expression level of thousands of genes to recognize the changes in expression level among different biological states. In this paper, superiority of MA is established over GA, simulated annealing (SA), and tabu search (TS), while selecting the genes in microarray data. Experiments on three well-known data sets, namely DLBCL, leukemia, and prostate cancer, exhibit that MA yields more promising results than classical GA, SA, and TS.