Based on Filter and Wrapper Feature Selection for Multi-class Cancer Classification

• Main Authors: Cheng-Hung Chang, 張丞宏
• Other Authors: Hsueh-Chi Shih
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/44804738134056972941

碩士 === 雲林科技大學 === 資訊管理系碩士班 === 98 === Feature selection method will affect access to multiple categories of classification accuracy, In this study, Filters feature selection methods combined with Wrapper, Using support vector machines for classification. Cancer microarray data with a low number of samples of high dimensional data gene, Therefore, the correct classification rate calculation because of the large number of characteristics leading to long computing time. To address this issue, Application of appropriate feature selection can not have the characteristics of classified information removed, Contribute to classification calculation, This study Filters feature selection methods combined with Wrapper, Filters for feature extraction in the use of BW ratio, Wrapper using genetic algorithms, Wrapper using genetic algorithms, Finally using support vector machine classifier, Using cross validation classification accuracy assessment. The results show, Use Filters Wrapper feature selection methods combined, can reduce the number of feature and increase the classification accuracy.