Abstract
Slope One algorithm widely used as a collaborative filtering recommendation algorithm based on the item, have characteristics of high recommendation accuracy and being easy to implement. However, without considering information like expert similarity and item characteristics, unimportant users and items have been included in this algorithm. It impacts the users' rating accuracy. Therefore, on the basis of the weighted Slope One algorithm, on one hand, experts and K-nearest neighbor users are comprehensively considered. On the other hand, item characteristics and common item information are integrated in this paper. The user's comprehensive rating to the item is predicted mainly by experts and item characteristics. The experimental results on two different datasets of MovieLens show that the improved algorithm could not only improve the prediction accuracy, but also effectively deal with the sparsity of the rating matrix.