User Interest Model Based On Social Network Research

With the rapid increase of information on the Internet, people are wasting more time in searching and retrieving contents that fit their needs, which encourages the appearance of personalized service. The aim of personalized service is to provide distinctive service such as retrieving service or recommendation service to a user which meets his interest. User preference model plays the key role of personalized service and determines the service quality. The only way to provide high quality service to a user is to access his interest and build a preference model that can be processed by computers. In this paper we introduce social network and community detection to the research of retrieving user preferences and study the methods of building user preference model and predicting further preference of a user.Community detection method is studied first in this paper. After studying the present method we propose a local algorithm for community detection. The algorithm is mostly based on local information of the network, which ensures the efficiency of the method. Also the way to divide vertices into communities guarantees the accuracy of the algorithm. Finally the algorithm divides a network into different communities with same type of vertices inside. The algorithm can be applied to searching for neighbors while computing user preference model, which is the main part of collaborative filtering recommendation technique.Based on present methods a hybrid approach of building user preference model which contains clustering techniques and collaborative filtering techniques is proposed. In the approach group preference is acquired by the method of social network, then combined with personal preference comes the final present preference model of one user.After studying the approach of building user preference model, we also discuss how to predict further preference of a user and present two methods based on time series analysis. One is to make predictions of a user based on his history records, which is proved to be effective by the experiments; the other is to locate a user's potential interest based on the preference of his neighbors, which is very leading and inspiring. We present several methods of building user preference model and predicting further user preference in this paper. These methods are used for different purposes. They can not only produce benefits for service providers, but also bring convenience to users, which creates a win-win situation.