A cancer classification method based on association rules

Gene expression data based cancer classification is of great importance to the computer aided diagnosis. In this paper, we propose a novel cancer selection method, AR-SVM. In AR-SVM, association rules are used as feature extraction approach to catch the non-linear relation among different genes, and support vector machine is used to classify the transformed gene expression data. The proposed method achieves both high classification accuracy and good biological interpretability. The experimental results on various gene expression datasets show that AR-SVM achieves the highest classification accuracy in comparison with existing gene expression classification methods.