基于最小风险贝叶斯涉密邮件统计分类算法  

A study on algorithm of secret-involved mails statistic classification based on least risk Bayesian

作　　者：邢莉[1] 喻建平[2] 

机构地区：[1]深圳大学经济学院,深圳518060 [2]深圳大学信息工程学院,深圳518060

出　　处：《深圳大学学报（理工版）》2008年第3期282-286,共5页Journal of Shenzhen University(Science and Engineering)

基　　金：国家高技术发展研究计划资助项目(2003AA142060);广东省自然科学基金重点资助项目(04106250);深圳市科技计划资助项目(200513)

摘　　要：提出一种适用于涉密邮件统计分类的最小风险贝叶斯算法,用以提高涉密邮件统计分类准确性及其效率.算法以最小风险为分类原则,根据邮件内容的已有训练集,可自适应学习出涉密邮件分类的风险值,快速实现敏感邮件的最小风险扫描.统计分析表明,该算法准确有效,具有较好的适应性.According to the accuracy and efficiency of the statistical classification methods about secret-involved mails, a least risk Bayesian algorithm was proposed. The algorithm was applicable to the statistical classification of secret-involved mails. The principle of the classification was to minimize the risk. With the training set of the e-mails, this algorithm could learn adaptively the minimized risk of secret-involved mails, and get the minimal risk by scanning the sensitive mails. Meanwhile, based on the minimal risk, a fast realization of Bayesian classification, which could further enhance efficiency of the algorithm, was proposed. Analysis and experiments show that the algorithm is accurate, efficient, and quite adaptive to the problems.

关 键 词：涉密邮件 统计分类 最小风险贝叶斯 风险自适应 电子邮件安全 

分 类 号：O212[理学—概率论与数理统计]