Hybrid Filtering Techniques for Group Recommender Systems: Analysis of Aggregation and Consensus Strategy
With the rise of digital platforms, Group Recommender System (GRS) is becoming an indispensable tool across industries such as entertainment, travel, dining, and social networking. This paper investigates hybrid filtering techniques in group recommender systems, specifically focusing on aggregation and consensus strategies that enhance group satisfaction and personalization. Group recommender systems, is essential for collective decision-making on digital platforms, as they encounter unique challenges, including balancing individual preferences to avoid "least misery" scenarios and ensuring user fairness. Thus, this paper first employed a detailed systematic review to identify important research gaps and utilizes the Movie Lens dataset, containing 100,000 ratings across 1682 movies, to evaluate a K-Nearest Neighbors-Weighted Term Frequency-Inverse Document Frequency (KNN-WTFIDF) based approach for managing group preferences. Cross-validation (k=5) was applied, ensuring balanced training and testing subsets. the analysis reveals gender-based rating patterns, occupation-based rating behavior, and insights into the most and least represented user groups. Findings include a general rating consistency across occupations, with slight variations favoring scientists and retirees, while farmers and unemployed individuals rated the lowest. This work demonstrates how hybrid filtering techniques can enhance satisfaction and fairness in group recommendations, laying a foundation for future research in adaptive group personalization models and robust consensus strategies
Download Paper