A Personalized Recommendation Algorithm Based On Area Distribution

With the rapid development of Internet in recent years, all kinds of invalid information has been appearing in people’s vision. Personalized recommendation system is a useful method to solve the problem. At the same time, individual habits, education background, locations, etc may bring a big difference in terms of what users prefer or expect. As a result, context should be taken into consideration when we do research on recommendation algorithms.In this paper, a personalized movie recommendation framework is proposed to study the issue that how area distribution affects the performance of recommendation. The framework is now capable of making recommendation to both normal and boundary users. All the above is based on a pyramid structure,which is used to decomposes the space, in each node, we store a collaborative filtering model and users’ location information(Postcode). The final recommendation list is calculated on the weighted score of all the partial results in related nodes.At last we validated our ideals by comparative experiments, and analyzed the result in detail.We show that considering user’s location information really can outperform the others. In particular, it can improve the satisfaction for boundary users.