Gene Selection And Cancer Classification Based On Optimization Algorithm And Support Vector Machine

The rapidly developing DNA microarray technology allows researchers to measure expression of thousands of genes data in a single experiment. These data has a high application value for understanding the pathogenesis, disease diagnosis and gene-level drug development. However, the microarray data usually contains thousands of genes with a small number of samples, which cause serious curse of dimensionality and deteriorates the diagnosis accuracy. Moreover, this gives rise to difficulty to a lot of classifiers. cut down the cost of medical diagnosis.The gene expression data set is always "few samples, high dimensionality". To solve this problem, this thesis proposed gene selection and cancer classification method based on optimize algorithm and support vector machine. There are three aspects of contribution. (1) Proposed the gene selection method based on genetic algorithm and support vector machine optimization, the method confirm the most informative gene subset and it avoid to some high rundent genes are selected. (2) Proposed the method based on particle swarm optimization and support vector machine optimization, the algorithm select the most informative gene subset and choose the optimal parameters of SVM simultaneously. (3) Proposed the gene selection method based on hybrid particle swarm optimization and genetic algorithm. The main idea of hybrid PSO/GA algorithm is to integrate the GA operators (selection, crossover, and mutation) into the binary PSO algorithm.Experimental results with the public dataset suggest that the proposed strategy not only can confirm the most informative gene subset for diagnosis, but also improve the classification accuracy.