Personalized Spam Filtering Method Based On Decision-Theoretic Rough Set

With the rapid development of network technology, E-mail gets rapid popularization because of its economic, convenient and fast. At the same time, we are living in an era of diversity. The independence, variability and diversity of people’s thoughts are increasingly prominent. Therefore, personalization is the development trend of E-mail technology in the future. At present, the study of personalized E-mail filtering technology is less. This paper apply multi-view decision model based on decision-theoretic rough set (DTRS) to the personalization of spam filtering. The users determine the loss value according to their own risk preference, and the E-mail server classifies the E-mail base on the loss value. The main work in this paper is summarized as follows:Firstly, correct the value range of a parameter in the existing multi-view decision model, and further complete the content of this model.Secondly, aiming at the calculation problem of the conditional probability of classification, this paper proposes a new estimation method based on vector space model (VSM). By taking into consideration the weight of feature words, the model is more reasonable.Thirdly, from the perspective of E-mail users, puts forward a set of value scale of the loss value, making it more easily for users to understand and to give desirable value.Fourthly, put forward the personalized E-mail filtering method based on DTRS. The E-mail server can classify E-mails into three classes according to three-way decisions or classify into two classes according to two-way decisions. The experimental results show that the proposed method can reflect the user’s personality and it can reduce the misjudgment rate, improve the classification accuracy when use the three-way decisions to classify.