基于用户限制聚类的协同过滤推荐算法

Collaborative filtering recommendation algorithm based on user restriction clustering

协同过滤推荐算法是个性化推荐系统的关键技术,但是其在计算过程中存在"数据稀疏性"和"冷启动"等问题,利用数据挖掘中的聚类技术,提出了一种基于用户限制聚类的协同过滤推荐算法。首先利用限制聚类技术将相似用户聚类到一起,在聚类的簇中找邻居用户,然后通过改进的协同过滤算法来进行推荐。实验结果表明,新算法改善了"数据稀疏性"和"冷启动"的问题,并且相比传统的协同过滤算法和基于K-means用户聚类协同过滤算法具有较高的推荐质量。

Collaborative filtering recommendation algorithm is the key technology of personalized recommender systems, but there exist some problems such as ＇ data sparseness＇ and ＇ cold start＇ in the calculation process. A collaborative filtering recommendation algorithm based on user restriction clustering is proposed by using the clustering technique in data mining. Firstly, the similar users are clustered together by the restriction clustering technique, and the neighbor users are found in the clustering. Then, the improved collaborative filtering algorithm is used to make recommendations. Experimental results show that the algorithm improves the situation of ＇ data sparseness＇ and ＇ cold start＇ , and has higher recommendation quality than the traditional collaborative filtering algorithm and the K-means user clustering collaborative filtering algorithm.