聚类加权和CS-LSSVM的文本分类  

Text categorization model based on clustering weighted and least square support vector machine optimized by cuckoo search algorithm

作　　者：赵专政[1] 李云翔[1] 

机构地区：[1]湖南城市学院数学与计算科学学院,湖南益阳413000

出　　处：《计算机工程与应用》2013年第16期124-128,共5页Computer Engineering and Applications

基　　金：湖南省教育厅青年项目(No.12B04)

摘　　要：文本特征提取和分类器优化是文本分类的两个关键问题,为了提高文本分类正确率,提出一种聚类加权(CW)和布谷鸟(CS)算法优化最小二乘支持向量机(LSSVM)的文本分类模型。采用TF-IDF算法计算特征词的权重,根据特征词的位置进行加权,经过特征聚类处理降低特征冗余度,采用LSSVM建立文本分类器,采用CS算法对LSSVM参数进行优化。采用复旦大学语料库对模型性能进行仿真测试,仿真结果表明,模型不仅提高了文本分类的正确率,而且提高了文本分类的效率。Text feature extraction and classifier optimization are two key problems for text categorization, in order to improve correct rate of text classification, this paper proposes a text classification model based on Clustering Weighted （CW） and Least Square Support Vector Machine （LSSVM） optimized by the Cuckoo Search （CS） algorithm. TF-IDF algorithm is used to calcu- late the feature weights, the feature is weighted by words position and features are clustered to reduced feature redundancy, the LSSVM is used to build text classifier which is optimized by CS algorithm. Fudan University data is used to test the perfor- mance of the proposed model. The simulation results show that the proposed model not only improves the classification accura- cy, but also improves the efficiency of text classification.

关 键 词：文本特征 聚类加权 最小二乘支持向量机 布谷鸟搜索算法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]