According to the problems such as cold start,sparse data existed in the traditional collaborative filtering algorithms,a hybrid collaborative filtering algorithm is proposed which combines user-based and item-based collaborative filtering.An improved algorithm is proposed to improve the accuracy of similarity calculation in the similarity algorithm.The control factors and balance factors are intro-duced in the missing data prediction process for the finally comprehensive recommendation.MovieLens dataset is applied in the experi-ments,the mean absolute error is used for the experiment as a test standard.Experimental results show that the user-item hybrid collabora-tive filtering algorithm can improve the recommendation performance and prediction accuracy in the extremely sparse matrix.%针对传统协同过滤方法中存在的冷启动和数据稀疏等问题,结合基于用户的协同过滤和基于项目的协同过滤提出一种混合协同过滤算法。在相似度的计算中提出改进算法来提高相似度计算的精确度;在预测未评分值时引入控制因子、平衡因子进行加权综合预测,最后再进行综合推荐。实验过程中采用MovieLens数据集作为测试数据,同时采用平均绝对误差作为实验的测试标准。实验结果表明,基于用户-项目混合协同过滤算法在评分矩阵极度稀疏的环境下提高了推荐的性能,并能有效提高预测的精度。
展开▼