The Application Of Double-index-Based SIS Variable Selection In Classification Problem

This paper devotes to variable selection and classification for microarray data,namely Leukemia 72 data.It uses a double index method,which is the combination of Sure Independent Screening()and importance priority method,to the exploration of variable selection of classification data.Hypothesis test are exploited to prove that the null hypothesis proposed by Robert Tibshirani in paper[15] is unreasonable.Then,we modified (94)6)and proposed a new statistic (94)6).By using the new (94)6)and method,we proposed three new variable selection models(called model I,model II,model III,respectively)for each data type and statistic.Since ignores correlation between variables,a new method- is proposed by adding importance priority.At the same time,(1 , and are utilized as classifiers after the procedure of variable selection,and the best classification model is given according to the Classification Error Rate()criterion.The results of numerical simulations and real example data illustrate the superiority of our proposed method,by the comparison with the rank sum test and fast screening variables method.