Vol. 2 No. 02 (2025): AI-Driven Hybrid System for Personalized Movie Recommendations

Abstract:

This paper introduces a novel hybrid recommendation system that integrates content-based and collaborative filtering approaches using deep learning techniques to enhance movie recommendations. Our model merges the metadata of movies, including genres, cast, and crew from the MovieLens dataset with user ratings to construct a comprehensive feature set. We employ a Term Frequency-Inverse Document Frequency (TF-IDF) vectorizer to extract content-based features and utilize Singular Value Decomposition (SVD) to derive collaborative filtering features, thereby addressing both user preferences and item characteristics.We further enhance the model by concatenating these features into a unified representation, which is then processed through a deep neural network to predict movie ratings. The network architecture consists of multiple dense layers with dropout regularization to prevent overfitting, ensuring robustness in learning complex user-item interactions. We evaluate our model on a standard dataset, focusing on mean squared error (MSE) as the performance metric to assess accuracy.The results demonstrate the effectiveness of our hybrid approach in providing precise recommendations by leveraging both the semantic content of movies and the historical interactions of users, thereby outperforming traditional methods that rely on singular recommendation strategies. This research contributes to the recommendation system community by showcasing a scalable and efficient method to improve recommendation quality and user satisfaction in multimedia services.

Keywords- Recommendation systems, Deep learning, Hybrid models, Collaborative filtering, Content-based filtering, MovieLens dataset, TF-IDF vectorization, Singular Value Decomposition, Neural networks, User-item interaction, Multimedia services, Movie recommendations, Personalization, Machine learning, Artificial intelligence.