用户和相似性填充相融合的协同推荐模型

Collaborative recommendation model based on user and similarity filling

针对当前模型存在的精度低、冷启动等难题,设计用户和相似性填充相融合的协同过滤推荐模型。采用稀疏评分矩阵建立用户评分项目集合,对评分矩阵进行降维、构建近似评分矩阵,采用该矩阵填充用户项目的评分集合,得到用户间相似度,采用信任度因子对填充误差进行合理调整,得到最终的用户间相似度,采用标准数据集Movielens和BookCrossing对模型性能进行测试。测试结果表明,该模型获得了较高的整体预测精度,预测准确性优于其它模型,改善了协同过滤推荐效果。

Aiming at the problems of low precision and cold start in the current collaborative filtering recommendation model,a collaborative filtering recommendation model based on user and similarity filling was designed.User score items were set up using sparse score matrix,and score matrix was reduced and approximate score matrix was constructed while matrix was used to fill user item score set to get similarity between users,Trust factor was used to adjust filling error and obtain final user similarity.The performance of the model was tested using Movielens and Book-Crossing dataset.The results show that collaborative filtering recommendation model gets high prediction accuracy,and prediction accuracy is higher than other collaborative filtering recommendation models,and recommendation effect of system is improved.