基于支持向量机算法和其他算法在文本分类中的性能比较  

Performance Comparison Based on Support Vector Machine Algorithm and Other Algorithms in Text Categorization

作　　者：木林[1] 

机构地区：[1]呼伦贝尔学院计算机科学与技术学院,内蒙古呼伦贝尔021008

出　　处：《内蒙古大学学报（自然科学版）》2011年第6期703-707,共5页Journal of Inner Mongolia University：Natural Science Edition

基　　金：内蒙古自治区高等学校科学研究项目(NJZZ11230)

摘　　要：分析了几种典型的文本分类算法的特点.其中,支持向量机算法具有最优性能,但处理大量数据时,耗时比较大;K-NN方法与K值的选取有很大关系,而且实时性不够良好;朴素贝叶斯方法在实际应用中可以得到比较好的效果;神经网络方法提供了比较容易的方式预测非线性系统,训练过程很慢,不能适应大数据量的学习.Characteristics of several typical text categorization algorithms are analyzed.Among these algorithms,the support vector machine algorithm has the optimal performance,but when large amounts of data to be processed,it takes a long time;K-NN method has much to do with the selection of K value,and its real-time performance is not good enough;Naive Bayes method can get better results in practice;Neural network method provides a relatively easy way to predict nonlinear systems,but the training process is very slow,which can not meet the learning with a large amount of data.

关 键 词：支持向量机方法 K-NN方法 朴素贝叶斯方法 神经网络方法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]