Advance Missing Data Processing for Collaborative Filtering

Memory-based collaborative filtering (CF) is widely used in the recommendation system based on the similar users or items. But all of these approaches suffer from data sparsity. In many cases, the user- item matrix is quite sparse, which directly leads to inaccurate recommend results. This paper focuses the memory-based collaborative filtering problem on the factor: missing data processing. We propose an advance missing data processing includes two steps: (1) using enhanced CHARM algorithm for mining closed subsets - group of users that share interest in some items, (2) using adjusted Slope One algorithm base on subsets for utilizing not only information of both users and items but also information that fall neither in the user array nor in the item array. After that, we use Pearson Correlation Coefficient algorithm for predicting rating for active user. Finally, the empirical evaluation results reveal that the proposed approach outperforms other state-of-the-art CF algorithms and it is more robust against data sparsity.