A Novel Approach to Movie Recommendation Using Weighted Collaborative Filtering with Activity and Rating Variability Analysis

Abstract

This paper introduces a novel approach to movie recommendation systems through a weighted collaborative filtering technique that integrates user activity and rating variability analysis. Traditional methods, such as Pearson correlation, fail to account for variations in user activity and rating behaviours. To address this gap, we propose a weighted Pearson correlation that adjusts similarity scores based on both the number of ratings and the variability in those ratings between users. This adjustment improves the precision and robustness of the recommendations. The similarity calculation is further refined by incorporating weights that reflect user activity and rating consistency, which are essential in mitigating biases introduced by users with differing rating patterns. Furthermore, the methodology incorporates a weighted adjustment formula to improve the prediction of user ratings for unrated items. Experimental results show that the proposed algorithm surpasses traditional methods, delivering better prediction accuracy and higher recommendation quality. The findings emphasize the effectiveness of incorporating activity and rating variability into collaborative filtering, resulting in a more reliable and robust recommendation system for real-world applications.