Reserch On Context Aware Personalied Recommendation Algorithm

Recommended system, which is an important means for information filtering, is a very promising method to solve the imformation overload. By reserching users’ interest preferences and conducting personalized calculation, recommended system can find users’s interest points and guide users find their own information needs. Currently, recommended system has been widely used in many fields. In the reserch field of recommended systems, traditional recommendation algorithms often only consider the similarity relationship between the user and the item, but less consider the specific context in which the item will be consumed, for example, time, location, season, weather, the surrounding people, emotion and so on. However, in many application scenarios, only relying on the binary relationship between the user and the item does not generate an effective recommendation. In the specific context, in order to recommend items that best meet user’s interest to the target user, the context where the item will be consumed is also very important for improving the performance of recommended systems. Therefore, accurately understanding the context where users and items are and incorporating the relevant context into recommendation algorithms are key steps to design a good recommended system. Context-aware recommended systems can provide more accurate rating predictions and more relevant recommendation by exploiting additional contextual information. Context-aware recommended system has been gradually becoming a very active branch of the recommended system reserch field.In this paper, we mainly reserch the context aware recommendation algorithms in the field of context aware recommended system. Firstly, we reserch the existing two kinds context aware recommendation algorithms based on the matrix factorization technology, and summary their advantages and disadvantages; then, to amend deficiencies of the existing context aware recommendation algorithms based on matrix factorization, we propose two improved context aware recommendation algorithms based on matrix factorization; finally, we conduct comparative expriments for the two improved algorithms and the existing two kinds context aware recommendation algorithms. Experimental results show that the two improved algorithms proposed in this paper outperform the exsiting two kinds context aware recommendation algorithms in the prediction performance.