Skip to Main Content
Microarray technology allows to measure the expression of thousands of genes simultaneously, and under tens of specific conditions. But the complexity of the data it produced makes it difficult to analyse. So reducing its high dimensionality is useful for both visualization and further clustering of samples. Traditional feature selection methods have various deficiencies. In this paper we propose two novel algorithms based on energy and maximum eigenvalue for feature selection, and test it on the leukemia dataset. we explore the use of support vector machines (SVM) for classification in the microarray analysis. The favorable results we obtained show that our methods for feature selection outperform other methods. At last, Analysis of our results with ROC also shows that our approaches for feature selection perform well.
Date of Conference: 23-26 Sept. 2010