协同过滤中基于用户兴趣度的相似性度量方法

在个性化推荐算法中，相似性计算方法是决定算法推荐效率的关键。通过分析传统的相似性度量方法在推荐系统中存在的不足，提出了一种基于用户兴趣度的相似性计算方法。该方法利用用户对不同项目类别的兴趣程度与用户评分相结合进行用户之间的相似性计算，克服了传统相似性计算方法仅仅依据用户评分进行相似性计算的不足，并在一定程度上减少了评价数据稀疏的负面影响。实验结果表明，该方法可以有效地克服传统相似性方法中存在的不足，使推荐系统的推荐质量有明显提高。

Similarity measurement based on user interest in collaborative filtering

In the recommendation algorithm, similarity measurement is fundamental to the recommendatory effectiveness. Through analyzing the problems of traditional similarity measurement in recommendation system, a new interest-based similarity measure approach was proposed, which used user degree of interest in different kinds of item with rating of user to calculate similarity score between two users, so that could overcome the drawback of only using rating of user to calculate similarity on traditional similarity measurement and overcome effect of extreme sparsity of user rating data. The experimental results show that this method can effectively solve the shortcomings of traditional similarity method, and provide better recommendation results than traditional similarity measurement.