一种基于情景的协同过滤推荐算法

针对协同过滤推荐算法中数据极端稀疏所带来的推荐精度低下的问题，文中提出一种基于情景的协同过滤推荐算法。通过引入项目情景相似度的概念，基于项目情景相似度改进了用户之间相似度的计算公式，并将此方法应用至用户离线聚类过程中，最终利用用户聚类矩阵和用户评分数据产生在线推荐。实验结果表明，该算法能够在数据稀疏的情况下定位目标用户的最近邻，一定程度上缓解数据极端稀疏性引起的问题，并减少系统在线推荐的时间。

A Collaborative Filtering Recommendation Algorithm Based on Scenario

The extremely sparse data in collaborative filtering recommendation algorithm often causes the decline of recommendation＆#39;s precision. In order to solve the problem,suggest a new collaborative filtering recommendation algorithm based on scenario. By introducing the concept of item＆#39;s scenario-similarity,modify the formula of similarity between users is improved based on item scenario similarity, and then the method has been applied in the procedure of user-clustering in the offline phase and recommendation in the online phase is produced with user-clustering matrix and user evaluation data. The experimental results show that the algorithm can locate the nearest neighbor for target in the condition of sparse data,alleviating some sparsity problem and reducing the time of recommendation in the on-line phase.