Fusion Project Popularity Of Collaborative Filtering Recommendation Algorithm

With the rapid development of the internet information technology, “information overload” is getting increasingly significant. In order to find the product which the user interests, the recommended system is brought out. As the core of the recommended system, recommendation alogrithm has become a focus for researchers. But as a good system, it’s just not enough to recommend favourable products for users, but the fact that the popularity of the product will vary over time should be considered. Consequently, to avoid the popularity problem, the time factor needs to be taken into account in the designing process of the alogrithm.In this paper, the time factor is introduced in the project-based collaborative filtering algorithm. As the popularity of different projects is varied with time scale, it’s very necessary to consider the popularity in the recommended time. At first, the time window and the popularity of project at any time window are defined; Then using time series forecasting model to predict the project popularity in the recommended time window; Then introduce fusion factor, linear weighted method is used to combine the traditional project-based collaborative filtering algorithm and the popularity of the project combination; Finally the combined algorithm is proposed. In this alogrithm, the project popularity in the recommended time window is fully considered. When a new user join into the system, the user’s historical behavior data can not be obtained, the recommendation algorithm will recommend high popularity projects, the recommendation result will not cause lag.To verify the the proposed algorithm, the movie rating data is select as a data set. Iterative method with the time window is used in comparative experiments to verify the difference between the new algorithm and traditional algorithms. Optimal value obtained by experiment for the fusion factor is 0.12. The proposed fusion algorithm compared to the traditional algorithm, the recommendation effect is getting improved, the algorithm is superior to traditional methods.