Collaborative Filtering Recommendation Algorithm Based on Item Clustering and Global Similarity

Collaborative filtering is one of the most important algorithms applied in e-commerce recommendation systems. The conventional calculations of similarities are inefficient, which suffers from data sparsity and poor prediction quality problems. In order to overcome the limitations, a new collaborative filtering recommendation algorithm based on item clustering and global similarity is proposed. Firstly, K-MEANS clustering algorithm is applied to cluster items into several classes based on users´ ratings on items, and the local user similarity is calculated in each cluster. In addition, the factor of overlap is introduced to optimize the accuracy of the local similarity between users. Finally, a newly global similarity between users is presented to optimize the selection of target user´s neighbors and achieve better prediction accuracy. The experimental results show that this method can improve the accuracy of the prediction and enhance the recommendation quality.