利用云模型改进基于项目的协同过滤推荐算法

Applying Cloud Model to Improve Item-Based Collaborative Filtering Recommendation Algorithm

针对基于项目的协同过滤推荐算法在数据极度稀疏的情况下相似性度量不准、推荐质量低下的不足,借鉴基于云模型中的云相似性度量方法来实现基于知识层面的项目相似性度量,改进传统的基于项目的协同过滤推荐算法,并利用公开的实验数据进行验证比较,结果表明,即使在数据极度稀疏的情况下,改进后的算法仍然能取得较好的推荐效果。

In accordance with the problem of item-based collaborative filtering （CF） algorithms, that on the measuring method of items＇ similarity works poor because of the extreme sparsity of user rating data and makes the quality of recommendation system decreased dramatically, a novel similarity measuring method on knowledge level is proposed, its thoughts comes from the measurement method of cloud similarity. Then, based on the novel method, item-based CF algorithms are improved. Experiments results show that the algorithms can achieve better prediction accuracy even with extremely sparsity of data.