An Improved Slope One Algorithm Combining KNN Method Weighted by User Similarity

Data sparsity is a main factor affecting the prediction accuracy of collaborative filtering. Based on the simple linear regression model, Slope One algorithm aims to enhance the performance significantly by reducing the response time and maintenance, and overcoming the cold start issue. It uses rating data to do calculation without considering the similarity. In this paper, we proposed an improved algorithm by combining the dynamic k-nearest-neighborhood method and the user similarity generated by the weighted information entropy with Slope One algorithm. Especially, the similarity between users is calculated and added on the fly. Experiments on the MovieLens data set show that the proposed algorithm can achieve better recommendation quality and prediction accuracy. Besides, the stability of the algorithm is also relatively satisfying.