摘要： 研究一种基于最小风险贝叶斯决策的垃圾短信过滤方法。对于以文本信息为主的短信，采用信息增益的方法进行特征选择，使用基于最小风险贝叶斯决策方法进行分类。通过自建短信语料库对该方法进行了实验。实验结果表明，该方法能够准确地对短信进行分类，降低合法短信的分类错误率，分类正确率达到99.3%，符合了短信分类要求。

关键词: 垃圾短信, 短信过滤, 文本分类, 朴素贝叶斯

Abstract: This paper analyzes a junk message filtering system based on the minimum risk Bayesian filtering algorithm, adopts Information Gain(IG) to select the feature, uses the minimum risk-based Bayesian filtering algorithm to classify. The experimental result, which data set is constructed by the real SMS from mobile company, shows that the method has a good performance on classification and low error rate on legit messages. The legit messages recall has achieve 99.3%. It’s suitable for SMS classification.

Key words: junk message, SMS filtering, text classification, naï, ve Bayesian

中图分类号: 

• TP309