论文部分内容阅读
为了提高垃圾邮件识别的准确度,减少识别中的错判,提出了一种交互式垃圾邮件识别方法。该方法用一组具有特定权重的规则识别垃圾邮件,规则权重分布用改进遗传算法训练得到。增加用户与服务器间的交互,收集用户反馈的错判信息,根据反馈信息用增量学习动态调整规则权重。通过对SpamA ssass in扩展实现了该方法,并应用在邮件服务器上进行了测试。实验中在不影响垃圾邮件识别率的前提下,降低误判率约10%。实验结果表明:该方法不但能有效减少识别中的误判,而且避免了繁琐的重新训练,加快了规则权重的更新速度。
In order to improve the accuracy of spam identification and reduce the misjudgment in identification, an interactive spam identification method is proposed. The method uses a set of rules with specific weight to identify spam, and the rule weight distribution is trained by improved genetic algorithm. Increase the interaction between users and servers, collect wrong feedback information from users, and dynamically adjust rule weight with incremental learning based on feedback information. This method is implemented by extending the SpamA ssass in and is tested on the mail server. Under the premise of not affecting the recognition rate of spam, the false positive rate is reduced by about 10%. Experimental results show that this method not only can effectively reduce the misjudgment in recognition, but also avoids cumbersome retraining and accelerates the update of rule weights.