A combining dimensionality reduction approach for cancer classification

Abstract

Because the original gene microarray data has many characteristics such as high dimension and big redundant, which is not good at classification and diagnosis of cancer. So it is very important to reduce the dimensionality and identify genes which contribute most to the classification of cancer. A method of dimensionality reduction based on the combination of mutual information and PCA is proposed in this paper. We adopted the SVM as the classifier in the experiment to evaluate the effectiveness of our method. The experimental results prove that the proposed method is an effective method for dimensionality reduction which can get very small subset of features and lead to a better classification performance.