A Synergistic Approach with Adaptive KNN and Matrix Factorization for Enhancing Movie Recommendations
Luong Vuong Nguyen, Quoc-Trinh Vo, Ho-Trong-Nguyen Pham, Thi-Thuy-Hoai Nguyen, Thi-Thu-Hong Phan · 2024
Improving movie recommendations is vital in personalized content delivery. We propose a novel approach that combines Adaptive K-nearest neighbors (KNN) and Matrix Factorization (MF) to tackle recommendation challenges. The Adaptive KNN method adjusts its neighborhood size based on user interactions, addressing data sparsity issues. It incorporates contextual information and user similarities to refine suggestions. Meanwhile, MF uncovers hidden patterns within user-item interactions, enhancing recommendation accuracy by understanding preferences more deeply. Using a MovieLens 25M dataset, we evaluated our approach. Comparative analyses against baseline methods confirmed our method’s superiority in accuracy, scalability, and adaptability. Cross-validation and sensitivity analysis affirmed its stability and versatility across diverse user preferences and dataset sizes. Our findings emphasize the synergy between Adaptive KNN and Matrix Factorization, surpassing the limitations of traditional systems. This work contributes to recommendation algorithm advancements, offering a promising framework for movie recommendations.