In recent years, many studies have shown the microarray gene expression data is useful for disease identification and cancer classification. Due to it only has small number of samples, and contains thousands of genes simultaneously, it leads difficulty to implement the classification studies. Previous researches have shown that not all of the genes are necessary for identification of cancer category. Therefore, to extract small numbers and relevant genes involved in different types of cancer is an urgent and essential assignment. In this paper, both of the filter and wrapper frameworks were used to embed in a new gene selection method. The proposed method was combined with K-nearest neighbor classified algorithm to evaluate the classification performance on six published cancer classification data sets. The experiment results showed that our proposed method could select fewer numbers of gene subsets and lead to better accuracy of predictions than other literature methods.
