Analysis on Gene Expression Data using Confidence Margin

Abstract

Due to recent advances in biotechnology, we can get activation level of each gene within an organism at a particular point of time. The data is called “gene expression data”. Analysis of gene expression data can provide understanding and insight into gene function and regulatory mechanism. However, these tasks are made more difficult from the empirical nature of array data and the overwhelming number of gene feature. One of our previous works in our field is GASVM. GASVM is a hybrid method of Genetic Algorithm and Support Vector Machine by Saberi et al.(1). GASVM has a large computational cost and a possibility of overfitting. Therefore, we have introduced a new criterion “Confidence Margin” and proposed a new method using it. The experimental result using two famous datasets confirmed the effectiveness of our proposed method.