A Technique of Recursive Reliability-Based Missing Data Imputation for Collaborative Filtering

Collaborative filtering (CF) is a recommendation technique that analyzes the behavior of various users and recommends the items preferred by users with similar preferences. However, CF methods suffer from poor recommendation accuracy when the user preference data used in the recommendation process is sparse. Data imputation can alleviate the data sparsity problem by substituting a virtual part of the missing user preferences. In this paper, we propose a k-recursive reliability-based imputation (k-RRI) that first selects data with high reliability and then recursively imputes data with additional selection while gradually lowering the reliability criterion. We also propose a new similarity measure that weights common interests and indifferences between users and items. The proposed method can overcome disregarding the importance of missing data and resolve the problem of poor data imputation of existing methods. The experimental results demonstrate that the proposed approach significantly improves recommendation accuracy compared to those resulting from the state-of-the-art methods while demanding less computational complexity.

Automated summary

A k-recursive reliability-based imputation method and a new similarity measure method are proposed in this paper to address the issue of poor recommendation accuracy in collaborative filtering methods when dealing with sparse data. Experimental results show that the proposed approach significantly improves recommendation accuracy compared to state-of-the-art methods, while also demanding less computational complexity.