In order to overcome the disadvantages of the restricted Boltzmann machine (RBM) for collaborative filtering algorithm,which ignored the fact that user's interests varied over time and only used the serious sparse user rating data.Firstly,this paper proposed a user-based RBM model with time information:TimeRBM,by adding the time information bias term into the existing RBM model.Secondly,it proposed a method of clustering based on the attributes of the item for rating prediction.Finally,it weighted and fused the two prediction results obtained by the TimeRBM model and the item attribute clustering,thus built an efficient hybrid algorithm.Experimental results on the benchmark dataset show that this hybrid algorithm can improve the prediction accuracy of the recommendation system.%针对受限波尔茨曼机用于协同过滤算法存在的不足,忽略了用户兴趣随时间变化,同时只利用了严重稀疏的用户评分数据,首先提出一种融合了时间信息的用户RBM模型:TimeRBM模型,即在原有RBM模型中加入时间偏置项;其次提出利用项目属性信息聚类的方法进行评分预测;最后将TimeRBM模型和项目属性聚类方法得到的两种预测结果进行加权融合得到一种高效的混合算法.在基准数据集上的实验结果表明,这种混合的算法有助于提高推荐系统的预测精度.
展开▼
