Optimal Genes Selection with a New Multi-objective Evolutional Algorithm Hybriding NSGA-II with EDA

Recent studies on molecular level classification of tissues with DNA microarray technology have produced remarkable results. It is believed that the subtypes of cancer can be distinguished by a set of discriminative genes. To achieve this goal, it not only requires high enough classification accuracy, but also a minimal number of genes as much as possible to lower cost. Meanwhile, the number of samples from different tissues may differ greatly. Therefore, it should also avoid classification bias due to unbalance sample number in different classes. In this paper, we propose a new multi-objective evolutional algorithm (MOEA) framework to select optimal genes, which has both advantages of the non-dominated sorting genetic algorithm II (NSGA-II) and the estimation of distribution algorithm (EDA). Finally, experiment on the data is done, The result shows that our method has good performance.