Abstract: When searching the nearest neighbor set, the traditional collaborative filtering algorithm ignores the impact of the time factor, only from the user or item takes into account the similarity of the user or item unilaterally, and ignores the impact of user characteristics on recommendation. Aiming to the above problems, we introduced the time forgotten function, resources viscosity function and the user feature vector, and improved the process of finding the user's nearest neighbor set, reflected the time effect, degree of user preferences and user characteristics, and tested this new methodology on data set got from MovicLens. I}hc results of experiment show that proposed method can improve the accuracy of the prediction.

Key words: Collaborative filtering, Time effect, Uscr preference degree, Uscr characteristic, Collaborative filtering,Recommendation system,Common grading items,Structural similarity,E-commerce