Weighted Slope One Algorithm Optimization Based on User Similarity and Item Similarity
Zuocheng Zhao, Jiachen Zhang · 2018
The traditional Slope One algorithm adopts the simplified regression form of linear regression to score unrated items. The principle of slope one algorithm is simple and easy to understand, with high query efficiency and reasonable accuracy, and supports online query and dynamic update, which makes them very suitable for the actual system. However, the Slope One algorithm and the weighted Slope One algorithm do not take into account the internal correlation between users and users, and between items and items. Using data of all users without distinction is likely to cause deviation and effect the recommendation quality. For the Slope One algorithm, user similarity and item similarity are not fully considered, this paper proposes an optimization method to improve its shortcomings. The improved algorithm adds user similarity and item similarity as a weight factor to the original formula. Experimental analysis of the MovieLens dataset shows that the optimized algorithm can improve the accuracy of two to three percent based on the original algorithm. In the case of sparse data sets and fewer neighbors, high recommendation accuracy and better convergence speed can still be achieved.