基于优化的多核学习方法的Web文本分类的研究  

Research of Web Document Classification Based on Optimized Multiple Kernel Learning Method

作　　者：江伟[1,2] 潘昊[1] 

机构地区：[1]武汉理工大学计算机科学与技术学院,湖北武汉430070 [2]武汉科技大学城市学院信息工程学部,湖北武汉430083

出　　处：《计算机技术与发展》2013年第10期80-82,86,共4页Computer Technology and Development

基　　金：湖北省自然科学基金(2011CDB257)

摘　　要：Web文本分类技术是数据挖掘中一个重要研究领域,为了能从海量信息中快速检索遍布网络各处的文档,需要提高Web文本分类技术的性能。多核学习方法是当前机器学习领域的一个热点,可以显著提升分类识别能力和学习推广能力,而核方法是解决高维非线性模式分析的有效方法之一。利用多核代替单核能增强决策函数的可解释性并获得更优的性能。文中分析研究了一种基于优化的多核学习的支持向量机,在此基础上结合通用的Web文本分类模型,提出了一种基于多核学习支持向量机的Web分类方法。通过实验测试表明,该方法具有良好的效果,对比一致组合的多核学习方法,所提出的方法具有较高的准确率。Web document classification has been considered as an important research field in data mining,it＇s necessary to improve the performance of technique of Web document classification for quickly retrieving the documents from the massive information spread all o-ver the network. Multiple-kernel learning is a focus in current machine learning community,which is able to develop the capability of classification and learning extension,while kernel method is one of effective approaches for solving high dimension and non-linear pattern analysis. By using the advantage of multiple kernel can boost interpretability of decision function and obtain better performance. In this paper,propose a Web document classification based on multiple kernel learning after a research of a SVM based on multiple kernel learn-ing. According to the result of the experiment,this approach presented in this paper has high efficiency and more accurate rate compared with simple consistent combination multiple kernel learning method.

关 键 词：支持向量机 数据挖掘 多核学习 WEB文本分类 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]