A 2-stages feature selection framework for colon cancer classification using SVM

As the colon cancer gene expression dataset is of high dimension, many irrelevant, redundant and noisy features might be included which may cause unprecedented challenges for data mining and machine learning algorithms. In this paper, we have proposed a new feature selection based method for colon cancer classification. First, we have used the ReliefF filter technique to provide a ranking in terms of the discriminatory ability of each feature. Second, since ReliefF cannot handle feature redundancy as well as feature interaction, another step is performed to select the best subset of gene expression profiles from the available 2K subsets. The proposed method has efficiently reduced the dimensionality of the colon dataset and increased the classification accuracy. The results from the Colon Cancer Gene Expression Data Set confirmed the effectiveness of the proposed method compared to advanced techniques.