SMS Classification Based On Association Classification

The Short Messaging Service(SMS) is an important function of a mobile phone. It provides a convenient and fast way for users to communicate with others. But at the same time, it also brings many troubles. A number of spam SMSs, such as illegal SMSs and advertising SMSs, are mixed in normal SMSs. Those spam SMSs disrupt people’s daily life seriously, even cause huge economic losses. Although scholars have made a lot of researches on this problem, the accuracy and recall of intercepting spam SMSs still need to be enhanced. Therefore, it is also a research hotspot to increase the accuracy and recall of intercepting spam SMSs effectively, which can improve the satisfaction of customers.The main works of this paper are:(1) Proposing ACW(Associative C lassification based on Word Order) algorithm. It improves the Apriori algorithm by making associative classification and word order together, adding word order into frequent item, which can make the classify rule more easily to be understood and adjusted.(2) Selecting the classification rules which satisfy the two conditions below. The one is having a higher accuracy than threshold, the other is that there is no including relation between the two correctly classified SMSs sets by any two rules.(3) Pretreating SMSs before mining classification rules. This includes word segmentation, clustering, replacing data, processing sensitive words, removing stop words and getting the feature words. By using these process, the pretreating process can improve the efficiency of generation, reduce the vector space of feature word s, reduce overhead of system.(4) The research achievement has met the design requirements by quantities of tests, and applied in the project of "Spam SMSs interception" in Lenovo Research Institute., which can prove the practical application value of this research.