A New Classification and Prediction Model With Two Stage Gene Selection Method Using Minimal Subsets of Gene Expression Data

Data mining models are extensively used in the field of disease diagnosis. Gene expression data are a main factor for the success of disease diagnosis. With thousands of gene expression data, gene selection is being a big challenge prior to classification. The proposed method incorporates two stages in gene selection. In the first stage pair wise gene selection was performed using a popular statistical technique. In the second stage the gene pairs that achieved 100% Cross Validation (CV) accuracy of those genes selected in first stage were used for classification. The testing results were compared with the single stage method and improvement on the computational burden was also proven to be the best in the proposed two-stage method. The paper also compares the performances of the three different classifiers Support Vector Machines (SVM), K Nearest Neighbour (KNN), Linear Discriminant Analysis (LDA) and promising results have been achieved.