一种基于R-FSVM的文本分类算法

A Kind of Text Classification Algorithm Based on Robust Fuzzy Support Vector Machines

目前模糊支持向量机使用的隶属度很大程度上依赖于先验知识、后验概率和多个自由参数,应用面不广、计算复杂、较难执行性能调优。为独立于学科领域知识,让模糊支持向量机利用样本集构造出性能更优的分类器,引入了一种更具鲁棒性的隶属度计算方法:R-FSVM,该算法使用重构误差计算每个文档相对于归属类别的隶属度。最后构造了文本分类实验系统,实验结果表明改进算法在准确率、召回率、F1值上具有更好的性能。

The fuzzy support vector machines using membership is more dependent on prior knowledge,posteriori probability and some free parameters.It makes application is not wide,computational complexity.In order to be more independent of subject area knowledge,fuzzy support vector machine use the sample sets to construct better performance of the classifier.It introduces a kind of more robust calcu lation method of the membership(R-FSVM),using the reconstruction error to calculate membership of Each document.Finally,it con struct a text classification experiment,the experimental results show that the improved algorithm has better performance in accuracy,recall rate,the F1 value.