基于聚类和项目相似性的Slope One算法优化

随着用户项目数量的增长，用户项目矩阵变得越来越稀疏，使用基于最小生成树的k-means算法对项目进行聚类并以聚类结果对用户评分矩阵进行预测填充。考虑到Slope One算法存在用户兴趣变化问题，将时间权重加入Slope One算法中进行评分预测。将改进后的算法在Movie Lens数据集上进行验证，结果表明，改进后的算法可有效解决稀疏性问题和用户兴趣变化问题，并将MAE值降低到0.015以下。

Optimization of Slope One Algorithm Based on Clustering and Project Similarity

With the growth of number of users, the user item matrix becomes more and more sparse. K-means algorithm based on minimum spanning tree of the project is used to cluster the items. The clustered results for the user rating matrix are used to predict filling. Taking into account the user interest change in Slope One algorithm, time weight added to the Slope One algorithm is used to predict rating. Experiments on the Movie Lens dataset show that the improved algorithm effectively solves the sparse and user interest change problem and the MAE value is reduced to less than 0.015.