Abstract: Abstract When facing with the extreme sparsity of user rating data, traditional similarity measure method performs poor work which results in poor recommendation duality. To address the matter, a new collaborative filtering recommen- elation algorithm based on item clustering and global nearest neighbor set was proposed. Clustering algorithm is applied to cluster items into several classes based on the similarity of the items, and then the local user similarity is calculated in each cluster, at last a newly global similarity between nearest neighbor users is used to measure user similarity. In addi- tion,the factor of overlap is introduced to optimize the accuracy of the local similarity between users. The experimental results show that this algorithm can improve the accuracy of the prediction and enhance the recommendation quality, which shows good result on the condition of the extreme sparse data.

Key words: Recommendation systems, Collaborative filtering, Clustering, Ulobe similarity, Overlap