The paper summarizes the current gene chip data feature selection methods and their efficacies, compares their advantages and disadvantages. It points out that the feature selection method which starts from its own characteristics can obtain better classification performance and biomedical applications, at the same time provide important foundations for finding disease-causing genes.