Maximum correlation minimum redundancy in weighted gene selection

Microarray technology has been recently used to analyze the behavior of thousands of genes simultaneously, and have an important role in diagnosis, detection and treatment methods. Reducing the size of the attributes (genes) with high potential for classification of microarray data analysis is thus an important goal. In this paper, we propose a new feature selection method based on maximum correlation and minimum redundancy (MCMR). In addition, a new method for weighting the genes has been introduced to select a final set of genes within all participated genes in cross validation procedure. The performance of proposed have been analyzed on two microarray data sets: colon cancer and breast cancer dataset. The results show that MCMR can increase the classification accuracy as well as reducing the number of selected genes significantly, compare to some other gene selection methods such as SNR (signal to noise ratio), PCC (Pearson Correlation Coefficient) and Fisher score.