Hybrid recommendation method IN sparse datasets: Combining content analysis and collaborative filtering

In the actual B2C e-commerce systems, the user-based collaborate filtering algorithm is one of the most common recommended approaches, but the item sparsity of the users' common evaluation influences the application of this algorithm, and the requests of the recommendation systems are not just accuracy, but also novelty. In this paper, in order to describe the users' interests and preferences more accurately and reduce the data sparsity, we considered the user's level of consumption on the basis, using the association rule mining formalized the competitive relationship between goods; using the time-based Bayesian probability formalize the complementary relationship between commodities, and through these relationship between the two commodities matches the users' requiring preferences and price preferences into the item sets of user evaluation. Simultaneously, in order to describe the effects of the numbers of common evaluations on computing the similarity of neighbors, the Pearson similar neighbor calculation method is improved. Finally, through the comparative experimental analysis based on the F1 method and the diversity measurement method, the algorithm significantly enhance the recommendation accuracy and recommendation novelty. All data were collected from the Jingdong Mall site.