Abstract: Gene expression profiles is a high-throughput data.However,only a small number of gene mutations related to tumor development.So,it is a huge challenge that design good algorithms to discover information Genes from microarray data.In this paper,we presented a data mining method named TGASVM (Test Genetic Algorithms Support Vector Machine),which as little as possible to elect information genes ,however,which have a good classification ability based on SVM.Compared with other similar algorithms,both classification of TCGASVM the accuracy and the number of information genes of TCGASVM are better.