Hybrid Movie Recommendation System Using User Partitioning and Log-Likelihood Based Content Comparison

Digital streaming platforms now host vast collections of movies, making it increasingly difficult for users to discover content aligned with their personal preferences. Traditional recommendation systems often struggle with challenges such as sparse user–item interactions, cold-start situations, and limited semantic understanding of movie descriptions. This paper presents a Hybrid Movie Recommendation System that integrates matrix factorization using ALS, demographic-based user partitioning, and content similarity powered by Log-Likelihood scoring and CBOW embeddings. The hybrid architecture intelligently merges latent collaborative patterns, demographic signals, and semantic content features to deliver accurate and personalized movie suggestions. Comprehensive evaluation results demonstrate significant improvements in prediction accuracy, recommendation stability, and scalability when compared to conventional collaborative or content-based methods. The proposed approach shows strong potential to enhance user engagement, enrich movie discovery experiences, and support intelligent recommendation pipelines in modern streaming platforms.