关键词: 垃圾博客过滤，词频特征，自相似特征，组合特征，贝叶斯分类

Abstract: Splog filtering has become a new hot area in the international in recent years. Most of the traditional filtering algorithms arc based on word frequency feature classification, which is quite redundancy and lack of relevance. According to this problem,a dynamic filtering algorithm based on the combination of features for splog(CFDSI))was proposed to solve the problem of low relevance and redundancy. The CFDSD algorithm uses self-similarity feathers and the attributes of author, at the same time adopts the 13ayesian classification algorithm to optimize word frequency feature classification. Experiments show that the algorithm is adaptable to dynamical updated features of the blog with time changes, and improves filtering efficiency, while reducing the time to filter splog.

Key words: Splog filtering, Term frequency features, Sclf-similarity features, Combined features, Bayesian classification