Minimum Gene Selection using BSWFM

摘要

In this paper, we propose a supervised gene selection method to classify tumor and normal samples based on the Bounded Sum of Weighted Fuzzy Membership Functions (BSWFM). This study compares the performance of a Neural Network with a Weighted Fuzzy Membership Function (NEWFM) with and without the proposed gene selection method. The superiority of the NEWFM with gene selection over the one without gene selection was demonstrated using a colon cancer dataset. Two thousand genes were used as inputs for the NEWFM without gene selection, and these resulted in accuracy, specificity, and sensitivity of 79%, 59.1% and 90%, respectively. A minimum of 19 genes were used as inputs for the NEWFM with gene selection, and these resulted in accuracy, specificity, and sensitivity of 87.4%, 72.7% and 95%, respectively. The results show that the NEWFM with gene selection performed better than the one without gene selection.