Collaborative Filtering Recommendation Algorithm Combining Tag Relevance and Score Differences
Guangyang Shi, Wanjun Yu, Xiongmei Chen, Tianning Li · 2023
The aim of this study is to propose a collaborative filtering recommendation algorithm that addresses the issue of inaccurate user preference mining in traditional collaborative filtering algorithms. The algorithm combines tag correlation and score differences to achieve accurate recommendations. It incorporates Newton's law of cooling to model the attenuation of users' interest in tags over time, and calculates user interest based on tag correlations. Additionally, it considers the maximum difference in user ratings from historical records, taking into account the impact of these factors on item ratings. The algorithm also considers the rating time to capture recent user preferences. Finally, it performs a weighted fusion of user label interest similarity and rating difference similarity to identify the most similar neighbors and make item predictions, scores, and recommendations. Experimental results demonstrate that this algorithm significantly improves recommendation accuracy.