两步选取基因策略.pdfVIP

Knowl Inf Syst (2011) 26:487–500 DOI 10.1007/s10115-010-0288-x REGULAR PAPER A two-stage gene selection scheme utilizing MRMR filter and GA wrapper Ali El Akadi · Aouatif Amine · Abdeljalil El Ouardighi · Driss Aboutajdine Received: 17 February 2009 / Revised: 4 January 2010 / Accepted: 16 January 2010 / Published online: 10 March 2010 © Springer-Verlag London Limited 2010 Abstract Gene expression data usually contain a large number of genes, but a small num- ber of samples. Feature selection for gene expression data aims at finding a set of genes that best discriminates biological samples of different types. In this paper, we propose a two-stage selection algorithm for genomic data by combining MRMR (Minimum Redundancy–Max- imum Relevance) and GA (Genetic Algorithm). In the first stage, MRMR is used to filter noisy and redundant genes in high-dimensional microarray data. In the second stage, the GA uses the classifier accuracy as a fitness function to select the highly discriminating genes. The proposed method is tested for tumor classification on five open datasets: NCI, Lymphoma, Lung, Leukemia and Colon using Support Vector Machine (SVM) and Naïve Bayes (NB) classifiers. The comparison of the MRMR-GA with MRMR filter and GA wrapper shows that our method is able to find the smallest gene subset that gives the most classification accuracy in leave-one-out cross-validation (LOOCV). Keywords Feature selection · Genetic algorithm · MRMR · Support Vector Machine · Naïve Bayes classifier · LOOCV 1 Introduction In recent years, the development of microarray technology has made it possible to analyze thousands or tens of thousands of genes simultaneously. However, the major problem in this analysis is the huge number of genes compared to the limited number of samples [35]. Most classification algorithms suffer from such a high-dimensional input space. Furthermore, most of the genes in arrays are irrelevant or redundant to some specifie