Mixture feature selection strategy applied in cancer classification from gene expression

Recently, cancer classification based on gene expression has been developed. This gives a hope for the discrimination of cancer to a more systematic direction. However, there´re many challenges existing in the new method. Maybe the most important one is the unbalance that so few training samples exist compared to so huge genes been collected. So feature selection becomes one center problem of the cancer classification. A novel mixture feature selection strategy has been proposed in this paper, it make use of the characters of filter and wrapper, and synthesis three feature selection methods: Pearson correlation analysis, Relief-F and SVM