基于mts-pca的特征选择方法分析及其在肿瘤分类中的应用feature selection method based on mts - pca analysis and its application in tumor classification.docx

Abstract
Biological Sciences puter Science is currently the fastest growing of the two disciplines. As these two disciplines, cross-product - Bioinformatics have played an important role in genome research. DNA microarray is one of a new research field in Bioinformatics. With the rapid development of gene chip technology, selecting genes to classify tumors has been widely applied. However, there are large number of redundant gene expression data, which constrains to obtain valuable classification information. These redundant irrelevant information will not only increase plexity of data processing, but also lower the quality of classification information. Feature selection in data mining removes redundant gene expression data and enhances the quality of the raw data so that data mining can obtain more valuable information, at the same time,the cost of calculation and the cost of analysis redundancy information are greatly reduced.
Support Vector Machines is based on VC dimension theory and minimum principle of structure risk in Statistical Learning Theory. For the situation of limited samples, SVM is seeking promise plexities of the model and learning ability, in the hope of obtaining the best Generalization Ability. SVM is paid attention to more and more by researchers by right of its good generalization ability and optimum solving, and the applications of SVM e more and more. This paper will take the SVM as the base leaner.
The following is a list of our major tasks in this thesis. (a) Proposed mts method by improving the traditional t-test method. To select important genes using mts method involves several steps. In the first step, a score based on mts method is calculated for each gene. In the second step, all the genes are rearranged according to their scores. The gene with the largest score is put in the first place of the ranking list, followed by the gene with the second largest score, and so on. Finally, only some top genes in the list are used for classi