基于特征选择的垃圾短信过滤研究

A Feature-selection-based Study on Junk-SMS-filtering

根据短信文本长度有限的特点,分别从句法、句式、语义等不同角度选取特征属性,然后设定属性权重,最后用平衡权重与大量实验总结得出的阈值相比较,实现短信文本分类。此方法计算量小,提高了分类速度和准确率,同时为实现用户过滤的个性化要求提供了方便的途径。

With the characteristics of the text length limitation of messages, characteristic properties are set from the perspective of syntax, grammar, sermntic and syntactic structure, Then property weight is set and finally, balance weight and threshold drawn from a large number of the experimental eondusions are compared to classify SMS. This method needs less calculation and accelerates classification, with more satisfactoly accuracy and easier way to filter junk SMS.