欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
如何高效地特征提取和分类算法设计是衡量基于内容邮件过滤技术优缺点的关键。针对互信息 MI（Mutual Information）特征提取算法和朴素贝叶斯分类算法，通过引入特征项区分度的概念，分析特征项在分类中区分能力之间的差异，进而提出一种兼顾特征项区分度和互信息的特征提取算法。通过进一步将区分度添加到分类算法设计中，最终提出一种加权朴素贝叶斯算法，高效地解决基于内容邮件过滤问题。实验结果证明，改进后的算法在召回率、精确率和正确率上均有明显提高，且分类性能更加稳定。
英文摘要：
How to efficiently extract the features and the classification algorithm design are two keys to measure the advantages and disad-vantages of content-based spam filtering technology.In allusion to mutual information （MI）feature extraction algorithm and nave Bayes clas-sification algorithm,and by introducing the concept of feature term discrimination （FTD）,we analyse the discrepancy of distinguishing ca-pacity of feature terms in categorising process,and then put forward a kind of feature extraction algorithm which gives the consideration to both FTD and MI.By further adding FTD to the design of classification algorithm,at last we present a weighted nave Bayes algorithm which solves the problem of content-base filtering efficiently.Experimental results show that the improved algorithm has significant improvement in terms of recall rate,precision rate and accuracy rate,and the performance of classification is more stable as well.
同期刊论文项目
同项目期刊论文
期刊信息
位置：