Advancing Diversity in Recommendation Systems Through Collaborative Filtering: A Focus on Media Content

A recommendation system provides suggestions based on user preferences, interests, or behavior. However, a major challenge is its tendency to generate monotonous recommendations, reducing diversity and limiting new user experiences. Therefore, increasing diversity is essential to enhance user experience and satisfaction while maintaining recommendation accuracy. This research proposes to apply collaborative filtering method, which focuses on item-based filtering using KNN. This method focuses on item similarity using cosine similarity. To enhance diversity, the system filters results based on similarity and rating thresholds. The evaluation results confirm that applying a similarity threshold increases recommendation diversity, as indicated by consistently higher individual diversity values. Clustering further enhances individual diversity. Findings show that the highest individual diversity with clustering reaches 0.5719, compared to 0.5706 without clustering. These improvements suggest potential applications in domains such as e-commerce and music recommendation systems.