Research On Determining Method Of Relationships Between Sms Contacts Based On Support Vector Machine

SMS(short message service) is a commonly used communication tool in people's daily life; it may be useful for SMS data reapplication to obtaining the relationship between contacts in reality by the analysis of its content. The current study of SMS contact relations determine work is not much, but the research is important both in theory and practicality, so it is necessary to expand their research work.Considering the chinese text messages are informal and unstructured, on the basis of the introduction and analysis for the common text segmentation methods, based on the characteristics of the text, choose an appropriate method for text segmentation operation. Besides, considering the segmentation results of the meaningless stop words, use the "stop list by Harbin Institute of Technology” to remove stop words from the segmentation results.The result of word segmentation is difficult to determine SMS contact relations directly. Therefore, it's necessary to choose the appropriate text representation method for abstraction and modeling, which facilitates to subsequent analysis. on the basis of the analysis of text representation to the traditional method, chose the representation of text vector space model(VSM), and according to the characteristics of text, design text feature vector to satisfy the practical requirements. And it's necessary to discuss how to calculate weights of component of feature vector. The aforementioned works provide the data foundation for subsequent work.In view of the needs of the classification work, defines the possible relationship between the contacts, considering that it's difficult to clear the characteristic mode for each kind of relationship in text vector space, choose the method of category to determining the relationship between SMS contacts. On the basis of the discussion and analysis on mainstream classification method, choose the support vector machine(SVM) method, considering the needs of determining SMS contacts relations, use the SVM decision tree model to construct the vector machine for multi-class classification. Besides, choose the appropriate kernel function to realize the nonlinear classification. On the basis of aforementioned work, the final SMS contact relationship classification decision algorithm is presented.At the end of the paper, a corresponding experiment scheme and the corresponding evaluation index is designed. After the experiment of determining SMS contacts relations, the results show that the method of determining message contact relations based on SVM is effective. The work of the comparison experiments show that the method based on SVM is improved in determining accuracy...