The Research Of Gene Expression Data Based On Support Vector Machine

Bioinformatics has been gotten great progress in recent years, with the rapid development of Gene Chip, it is become more efficient and reliable for getting gene expression data, and these data provide a foundation of uncovering the secret of life. They have the characteristics of high dimension, less sample, and far beyond the capacity and speed of traditional analytical method, so analytical on gene expression data has become the bottleneck of biological research, and it is increasingly urgent for processing, mining, analysis and understanding.SVM(Support Vector Machine) is a type of new machine learning method based on statistical learning theory. It uses the principle of minimizing structural risk, and preferable to be used to resolve small sample problem, especially for gene expression data with high dimension, small sample and nonlinear, SVM demonstrated a good performance.This paper mainly study classification of gene expression data based on SVM, including binary classification and multi-classification. This paper first introduces bioinformatics and the analytical of research status quo on gene expression data, and then expatiate classification of SVM and correlative theory in detail. Furthermore, aim to the actual situation ("the small number of samples, high-dimensions ") of gene expression data, this paper do feature selection on the dataset using signal-to-noise and recursive elimination method, and getting a subset of genes; and then using kernel principal component analysis and other feature selection method on the subset for reducing dimension, so ensuring in the situation of no loss of actual information, improving the accuracy of predication as possible; this paper puts forward two classification method as follows. Finally, this paper do contrastive experiment on several commonly used gene expression data, and the result shows that the accuracy of prediction gotten improved compares to traditional SVM classification method.