Model Selection of SVMs Using GA Approach

• Main Authors: Peng-Wei Chen, 陳芃暐
• Other Authors: Hahn-Ming Lee
• Format: Others
• Language: en_US
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/25448168557813361881

碩士 === 國立臺灣科技大學 === 電子工程系 === 91 === Support Vector Machines (SVMs) classification has became one of the promising and popular classification methods for various disciplines. It is based on the theory of structural risk minimization and has good generalization properties that have been demonstrated both theoretically and empirically. A major open problem in SVMs is model selection, which tunes the hyperparameters, kernel parameters and the slack penalty coefficient C. Model selection is usually done by minimizing either an estimate of generalization error or some other related performance measures. Due to the time-consuming reason of the common used grid method and k-fold cross-validation, many researchers make efforts in proposing computational inexpensive methods; however, they still have the difficulty to meet users’ requirements. Thus, in this thesis we propose a new model selection method, named GA-based model selection, to make the task of the method selection more efficiently. The proposed method introduces the superiority of genetic algorithm (GA), which is well known as a method for solving complex search problems in a large number of different disciplines, making the method not only automatic but also time reduced. We perform experiments on several benchmark datasets to compare the performance of the GA-based model selection with some current model selection methods and show that our method outperforms others in terms of computational time and accuracy. According to the results, the proposed method is helpful for users to do the model selection more efficiently.