以簡化粒子演算法為基礎之支持向量機分類法

A Support Vector Machine based on Simplified Swarm Optimization for Classification

張呈維(Cheng-Wei Chang)

指導教授 ： 葉維彰

國立清華大學/工學院/工業工程與工程管理學系/碩士(2012年)

摘要

在資料探勘的領域之中，資料分類往往是最常被討論的議題之一，已有許多研究方法透過建立數學模型用以區分未分類資料，而支持向量機(Support vector machine, SVM)則是近年以統計學習理論所延伸發展出來的機器學習方法，其已被廣泛地被應用於統計分類及回歸分析中。然而，利用支持向量機處理各種獨特的分類問題時，目前並無任何準則、規範供使用者參考以選擇具有較佳分類準確率之參數設定，通常使用者需要自行不斷測試各種參數組合或者將支持向量機結合一些具學習能力之演算法，以求得具有較佳準確率之參數組，進而將此參數組套入支持向量機以對未知資料進行分類。故本研究乃提出以簡化粒子演算法為基礎之支持向量機分類法(SSO-SVM)，本方法論乃以簡化粒子演算法(Simplified swarm optimization, SSO)作為產生支持向量機所需參數之基礎，以較簡單快速之方式產生具有較佳準確度之參數組合；而支持向量機之分類準確度則回饋予簡化粒子演算法，作為各參數組合所對應之適應度，以作為產生新參數組合之基礎。此外，本研究將利用UCI資料庫之多筆標竿問題進行實驗，以比較本方法論與其他相似之分類演算法於資料分類準確度上之優劣。

關鍵字

統計分類 ； 簡化粒子演算法 ； 支持向量機 ； 群體智能

並列摘要

In the field of data mining, classification is one of the most discussed issues that generating a generalized known structure to apply to new data. Recently, support vector machine (SVM) has been introduced for analyzing data and recognizing patterns. It’s a useful technique for data classification and regression analysis. However, while using SVM dealing with each unique classification problem; it is not known beforehand which parameter combination is the best for a given problem. Users often need to do random self-test or apply other algorithm to find an acceptable solution. In this study, we proposed a support vector machine classification combined with swarm intelligence algorithm, called Support Vector Machine based on Simplified Swarm Optimization (SSO-SVM). The simplified swarm optimization (SSO) is an emerging population-based stochastic optimization method, which belongs to both categories of swarm intelligence and evolutionary computation. In this paper, simplified swarm optimization (SSO) is used to implement a parameter combination selection, and support vector machine (SVM) serve as a fitness function of SSO for classification problem. The result indicates that the proposed SSO-SVM has better performance and more efficient than other method listed in this paper.

並列關鍵字

Classification ； Simplified Swarm Optimization ； Support Vector Machine ； Swarm Intelligence

參考文獻

[2] C. Cortes and V. Vapnik. Support-vector network. Machine Learning, 20:273-297, 1995.

[6] X. Jia, B. Zhao and Z. Chen, Genetic Algorithm Optimization Based on Support Vector Machine Image Interpolation, 2011 Cross Strait Quad-Regional Radio Science and Wireless Technology Conference (CSQRWC)

[7] W. Zhuo and L. Chu, The Algorithm of Text Classification Based on Rough Set and Support Vector Machine, 2010 2nd International Conference on Future Computer and Communication

[8] H. Pan, Y. Duan and L. Tan, Application for Web Text Categorization Based on Support Vector Machine, 2009 International Forum on Computer Science-Technology and Applications

[9] M. Zhou and Q. Wang, The On-line Electronic Commerce Forecast Based on Least Square Support Vector Machine, 2009 Second International Conference on Information and Computing Science