An improved recommendation algorithm via social behaviors

Recommendation is an important problem in the traditional field of data mining. As a consequence, various kinds of algorithms have been proposed in the last few years to improve the recommendation performance. However, many of them overlook users´ rating behaviors. In this paper, an improved recommendation algorithm with considering users´ habits and rating behaviors will be proposed. Firstly, calculate the ratings entropy for each user via users´ rating records which can reflect users´ behavioral features. Secondly, widely-used user-based method will be improved by considering rating entropy. Finally, the predicting rate is generated. Experimental results on the three datasets: Movie Lens, Netflix and RYM all suggest that the proposed method can enhance the algorithmic accuracy. Furthermore, since this method can improve those missing value more accurately via users´ similarity and rating behaviors which might shed some light on discovering users´ purchasing intention and optimizing the performance of recommender systems by human dynamics.