SVM Based Classification Mapping for User Navigation

In order to implement active service, a kind of user navigation based on Support Vector Machine (SVM) classification was proposed which solved many traditional problems such as lack of rules and adaptability and so on. Especially, this paper completed a dual classified coordination between user attribute classification and SVM user classification using mapping matrix. In our SVM based user navigation, we first regard User interest Set as input, then apply the SVM classifier to sort out users on the basis of their interests or attributes, and what is more, combine the merit that user attributes classification promise better service being offered to users whose using frequency is not high, a dual-classified collaborating navigation between user attribute classification, and SVM user classification by using mapping matrix is being proposed. We use the mapping method to mix their advantages of common classification methods which lead them to collaborate and to provide the new and old custom with navigation. Using Classification T to statistic the latest and the most frequently used demands of each user type, then applying a mapping from user attribute class (A) to user interest class (T) with the purpose to judge what kind of user type in Classification T has the most possibility the user might belong, to and then to provide the user with the latest services according to their interests. From the experiments, the fitting degree of A and T classification and the true personality of users are derived by tracking the demand of 50 users. We can see that with the increase of frequency in use, the matching degree between Users Attribute Classification (A) and real types of user are essentially the same, and User Personalized Classification (T) is gradually reaching saturation. Regarding the certain point as a dividing line, after using classification mapping, the user's true extent of matching has a broader applicability. SVM have excellent generalization performance and minimum structural risk, which is a good machine learning method that elicits objective regulation from complex statistical dates. So we have improved the user model of active service navigation system and user navigation combined with User Attribute Classification A and SVM User Interest Classification T by mapping matrix. Classification A and T using mapping matrix to bounding them can satisfy the requirement of a different user. The records in the Classified Statistic Library can provide feedback date and research information for further study between a personalized user and user attribute.