多類支向機之研究

A study of multiclass support vector machines

王景南(Jying-Nan Wang)

指導教授 ： 古思明 ； 龐金宗

元智大學/資訊學院/資訊管理學系/碩士(2003年)

摘要

支向機(Support Vector Machines, SVM)原本用於處理二類別分類問題，但在許多應用領域所需處理的問題為多類別分類問題；因此如何將二類支向機之觀念，有效延伸為多類支向機是目前研究的重要課題。本文先針對五種常見的多類支向機方法簡略介紹，再提出一種新的多類支向機：one-against-half method，並以實際資料庫測試，比較此方法和其它多類支向機處理分類問題的效果。經實驗結果得知，one-against-half method亦為處理多類別分類問題時，可以選用的分類方法之一。

關鍵字

支向機 ； 多類支向機 ； multi-classification ； One-against-half method

並列摘要

Support vector machines (SVM) was originally designed for binary classification. SVM has been recently applied to solve multi-class problems. And there lies the unsolving research issues on developing 2-class SVM into multi-class SVM. In this paper, five common multi-class SVMs have been reviewed and a new multi-class SVM "one-against-half method" has been proposed along with the comparison between the performance of one-against-half method and the other five multi-class SVMs. The experiments proved one- against-half method to be a qualified multi-class SVM.

並列關鍵字

One-against-half method ； Support Vector Machines ； multi-classification ； mutliclass SVM

參考文獻

Available at http://www.liacc.up.pt/ML/statlog/

[5] C.J.C. Burges. A tutorial on support vector machines for pattern recogition. Data Mining and Knowledge Discovery, 2(2):955-974,1998.

[6] C.-C. Chang and C.-J. Lin. LIBSVM: a library for support vector machines, 2001.

Software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm.

[7] C.-C. Chang, C.-W. Hsu, and C.-J. Lin. The analysis of decomposition methods for support vector machines. IEEE Transactions on Neural Networks 11 (4), 1003-1008. 2000.