摘要
针对个性化邮件过滤中接收的邮件没有规律、正常邮件和垃圾邮件存在严重类偏移等问题,提出一种改进的k最近邻(k-nearest neighbor algorithm,KNN)个性化邮件过滤方法。该方法主要是通过建立兴趣度模型(对兴趣度计算的改进,剔除用户习惯对建立兴趣度模型带来的影响)不断改变训练集,使得训练集中的文本始终代表用户最近的兴趣。然后通过对接收邮件的无规律和类偏移问题的研究,提出一种改进的KNN算法,该算法主要是对文本在聚类中的价值重新评定,使其对邮件文本进行了较好的分类。经实际验证,改进后的KNN个性化过滤方法能较好地实现对邮件进行个性化分类。
Regarding such traits as irregularity of receiving the emails and imbalance between email category and the quantity, the thesis puts forward a personalized improved KNN algorithm of filtering the emails. Improving the counting method on interests and removing the influence of the users' habits on the establishment of interest model, the method perfects KNN algorithm and thus achieves better classification of the emails based on the research as to the issues of category deviation. Experiments show that improved KNN filtering method could help the users classify the emails according to the users' interests much more accurately.
出处
《重庆邮电大学学报(自然科学版)》
北大核心
2011年第6期757-760,共4页
Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)
基金
国家自然科学基金(60773113)~~