Research On Technology Of Mining The Dynamically Interesting Profiles Of The Users

One of the most important technologies in the personalized information service system is to mine the personalized user profile. User profile mainly describes the characteristics and the relationships of different users. Since the quality of service is dominated by the user profile, building a high quality of user profile is necessary in the recommendation systems.At present, there are no unified standard of building user profile. Although the existing technologies of mining user profile own their respective advantages, there are two common weaknesses in these approaches. First, fail in finding the implying meanings in the sequences of the categories or the keywords which appear frequently in the user's log file. Second, almost all the traditional user interesting profiles are static which can not describe user interesting dynamically.This thesis mainly resolves two issues which can not be solved by the traditional technologies of mining user interesting profiles:First of all, the user's interesting profile is described by combining the user's interesting point group and the user's interesting vector group together. Just like the user interesting points, the directed sequences of the points indicate the user's interesting and behavior habits, the interesting points and the interesting vector are mined from the categories and the 2-category sets visited by the users frequently. Combing the interesting points and the interesting vectors, user interesting profile is described as a directed weighted graph. This kind of user interesting profile can describe the user's interests more accurately.For the second, this thesis proposed a algorithm of mining dynamic user interesting profile. Based on the least squares technique, this thesis proposed a method to mine the dynamic interesting profile by fitting the sequence of interesting points and interesting vectors, the dynamic user interesting profile can describe the process of changes of user's interesting.To verify the effectiveness of the approach proposed in this paper, personalized recommendation experiments are realized by using the interesting profile based on interesting points and interesting vectors and the dynamic user interesting profile in PPRS, respectively. The experiments result shows that the user profile based on interesting points and interesting vectors is quite available in content-based filtering recommendation system and the collaborative recommendation system, and the experiments result indicates that the dynamic user interesting profile can describe the user interesting more clearly than the static user profile.