For microarray data analysis, most of them focus on selecting relevant genes and calculating the classification accuracy by the selected relevant genes. This paper wants to detect the relation between the gene expression levels and the classes of a cancer (or a disease) to assist researchers for initial diagnosis. The proposed method is called a Two Stages Weighted Sampling strategy (TSWS strategy). According to the results, the performance of TSWS strategy is better than other existing methods in terms of the classification accuracy and the number of selected relevant genes. Furthermore, TSWS strategy also can use to understand and detect the relation between the gene expression levels and the classes of a cancer (or a disease).