Microarray data analysis can provide valuable information for cancer prediction and diagnosis. One of the challenges for microarray applications is to select an appropriate number of the most significant genes for data analysis. Besides, it is hard to accomplish a satisfactory classification results by using data mining techniques because of the dimensionality problem and the over-fitting problem. For this reason, it is desirable to select informative genes in order to improve classification accuracy of data mining algorithm. In this study, Singular Value Decomposition (SVD) is used to select informative genes and reduce the redundant information. Furthermore, information gain is used to determine useful features of data to get better classification performance. In the last step, the classification technique is applied to the selected features. We conducted some experimental work on subset of features from available datasets. The experimental results show that the proposed feature selection and dimension reduction gives better classification performance in terms of the area under the receiver operating characteristic curve (AUC) and the prediction accuracy.

No abstract