Research On Personalized Recommendation Algorithm Of Mobile Information Service Based On Multidimensional Context Information

With the rapid development and popularity of the mobile information service and the improvement of the technology of data acquisition, the information dimention we acquire increases rapidly. The traditional "user-item" recommendation system is relatively effective as the collected information scale is relatively small and the information dimention is relatively low. However, in the mobile information service environment, the "user-item" recommend method which is two-dimensional mode can’t work as effectively as before any more. It is because that the preference of users are effected by their particular situation and the personalize recommendation can’t work well enough without the context information. This article focuses on making full use of multidimensional context information to provide personalized recommendations to mobile users in mobile information service as effectively as possible.This paper analyzed and modeled multi-type context based on mobile information service environment and considering the high dimension situation information adequately in the mobile information service environment. The collaborative filtering recommendation algorithm based on the users is relatively mature in "user-item" two-dimensional recommendation algorithms. This paper combine this algorithm and multidimensional context information conditions by expanding the preference of users in single-dimentional context information to multi-dimentional context information, at the same time, improve the similarity calculation algorithm which is the most important algorithm in collaborative filtering recommendation algorithms. To overcome the difficulty of data sparsity problem based on the single-dimentional, as it’s an important step to expand to multi-dimentional preference prediction, the tensor decomposition method was introduced. Finally, this paper adopts the MovieLens public data sets which is famous on recommended area to do the algorithm simulation test. The result shows that the context information have an influence on the user preference by comparing with the traditional pattern of two-dimensional recommended recommendation algorithms. The result also shows more effective comparing with the recommendation method which integrate multi-dimentional context in another way. Through the study of the above content, this paper enriched the related theory in the field of personalized recommendation, provided both theoretical support and scientific basis for personalized recommendation in mobile information service research...