Extended Collaborative Filtering Approach for Genre-Based Movie Recommender Systems

Collaborative recommender systems (RSs) play a crucial role in predicting user behaviors and suggesting new items. However, existing RSs face challenges related to accuracy, scalability, and data sparsity. Often, user preferences can be inferred by categorizing their ratings into distinct classes based on specific categorical features. In this study, we propose a novel approach that leverages movie genres to discern users' preferences, classifying them into three categories: “hatred,” “medium,” and “loved” genres. Subsequently, we combine the user's genre preferences with a user similarity method based on Euclidean and Cosine distances to recommend new movies. The proposed method enhances the performance of RSs by incorporating genrebased preferences. The proposed method is evaluated using the MovieLens dataset. Experimental results demonstrate approximately 24%, 51% and 49% improvement in accuracy, precision and recall rates while explores other facets of recommendation performance, such as overall user satisfaction versus traditional kNN.