The Research Of Filtering Methods Of Spam Messages Based On SVM

With the popularity of cell phone use,the SMS has become the important way for people's communication .However, a large number of spam messages have also appeared at the same time which badly affects not only the people's normal life, but also the social stability and unity. How to filter spam messages technology on the block, which restricts the spread of spam messages has become an urgent and realistic issue.Combining with the characteristics of SMS and using the method of SVM; this article take the view form the content of the SMS and treat the problem of the spam messages as the problem of two kinds of identification problem (that is the process which to prove whether the SMS is spam messages or not),and finally vector machines spam filtering method was raised. This method combines the characteristics of SMS, content-based considerations, the use of support vector machine algorithm to classify text messages, and then complete the spam filter. The main work includes:1. Support vector machine is proposed based on the weakness of traditional filtering spam messages, such as the Classification accuracy is low, adaptive ability is poor and so on. And it describes the steps and methods to achieve the goal, and the involved key technologies were deeply analyzed which including characteristic dimension reduction way, text representation methods and Classification algorithm .what's more, the experiment was done to identify the penalty parameters and kernel function which adopt to SMS classification of the SVM.2. According to problems form the SMS classification in the standard SVM.method such as noise and lack of information, the improvement method was put forward .This method using the SVM to identify whether the Key features were involved in the SMS, and put the recognition results to the original feature space, and repeated handling characteristics and noise, and then to identify spam messages.3. According to the proposed text based on support vector machine classification method, the integration of traditional spam filtering technology to construct a simulation system spam filtering, and spam filtering methods were experimental comparisons. Experiments show that the proposed text based on support vector machine method is effective to improve the filtering spam filtering accuracy.