Adaptive Movie Recommendations: A Deep Learning Framework with Mood and Sentiment Integration

This research study proposes an advanced movie recommendation system that combines deep learning models with mood detection and sentiment analysis for highly personalized suggestions. The system collects data from diverse sources, including user watch history from OTT platforms along with real-time social media sentiment from social media platforms. Additionally, wearable devices track heart rate to infer the user's emotional state during media interaction. The recommendation engine uses a multi-layered approach. Neural Collaborative Filtering (NCF) analyzes watch patterns, while BERT-based embeddings capture key movie metadata such as genres and actors for content-based filtering. The system also employs an LSTM model for mood-based recommendations, integrating heart rate data and social media sentiment to align suggestions with the user's current mood. Key advantages include a hybrid recommendation approach that combines collaborative filtering, content-based suggestions, and real-time mood analysis. The system adapts to both long-term preferences and short-term emotional states, delivering more relevant and personalized results. Deployed on scalable cloud infrastructure (AWS EC2), with continuous learning from user feedback, the system ensures dynamic and evolving movie recommendations that enhance user satisfaction.