An Improved Collaborative Filtering Algorithm Based on User Interest Diffusion and Time Correlation

This paper proposes an improved collaborative filtering recommendation algorithm based on user interest diffusion and the time correlation. Firstly, the algorithm improves user synthesis similarity calculation method based on user interest diffusion, calculates the direct similarity of user interest and the similarity of user interest diffusion, and obtains the synthesis similarity of user interest through parameter adjustment. Then, for the user interest change with time, the time correlation function is applied to the similarity calculation between users. Finally, the recommendation weight is divided into the time correlation data weight and the synthesis similarity data weight, so that a more accurate prediction score is obtained.. The comparison experiments showed that the algorithm can reduce the sparseness of the data set effectively when the data is sparse, and improves the precision of the recommendation algorithm.