Two-Stage Gene Selection Algorithms for Classification of Gene Expression Data

• Main Authors: Chao-Hsuan Ke, 柯兆軒
• Other Authors: Cheng-Hong Yang
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/85395768415812242759

碩士 === 國立高雄應用科技大學 === 電子與資訊工程研究所碩士班 === 96 === The microarray is a medical diagnostic tool with good efficiency, and it was used for analyzing the behavior characteristic between the gene and disease by the extensive one at present. Microarray data are characterized by a high dimension, which could be analyzed more than thousand of genes and diseases simultaneously. However, it will lead to need more computation time when it is implemented on classification. Many previous literatures showed the feature (gene) selection has some advantage, such as gene extraction which influences classification accuracy effectively, to eliminate the useless genes and improve the calculation performance and classification accuracy. The goal of this study is to select a small set of genes which are useful to the classification task. We proposed a two-stage method using several filter methods to proceed gene ranking and combined the evolutional algorithms on gene expression data to select an optimal gene subset. In this study, an improved particle swarm optimization which introduced a Boolean function was used to improve the disadvantage of standard binary particle swarm optimization as a new evolutional algorithm for gene selection, and both k-nearest neighbor and support vector machine classifiers were used to calculate the classification accuracy. The experimental results revealed that our proposed feature selection method is able to effectively select the relevant gene subset and achieve better classification accuracy than the previous studies.