基于用户特征和相似置信度的协同过滤算法

Collaborative Filtering Algorithm Based on User Features and Similar Confidence

针对推荐系统中协同过滤算法存在的用户冷启动和数据稀疏性的问题,提出一种基于用户特征的相似度和基于置信度的相似度相融合的计算方法。该算法对用户的特征进行计算得到一个相似度,再考虑到能正常反映用户之间的相似兴趣而进行计算得到一个和置信度有关的相似度,将两个相似度的权重按“相加为1”的方式进行融合得到最终的相似度。实验结果表明,这一方法在数据较为稀疏、用户邻居数较少的情况下与传统的协同过滤算法相比有较好的推荐效果。

In order to solve the problems of user cold start and data sparsity in collaborative filtering algorithms in recommendation systems,a calculation method based on user features similarity and confidence-based similarity is proposed.The algorithm calculates the user's characteristics to obtain a similarity,and then calculates a similarity related to the confidence by considering the similarity that can normally reflect the similar interests between users.The weights of the two similarities are merged in the manner of"adding to 1"to obtain the final similarity.Experimental results show that this method has better recommendation effect than the traditional collaborative filtering algorithm when the data is sparse and the number of user neighbors is small.