A Multi-Source Approach to Film Recommendations using Social Media, Search Data, and Streaming History

In the rapidly evolving OTT platforms, providing highly personalized and relevant film recommendations has become a critical challenge. Traditional recommendation systems primarily rely on user ratings and watch history, neglecting the vast potential of other data sources such as social media interactions and search engine queries. This research presents a multi-source film recommendation framework, integrating data from social media posts, search behaviour, and OTT streaming history to deliver enhanced personalization. A novel algorithm, CoFiSent (Collaborative Filtering with Sentiment Analysis), is proposed, combining collaborative filtering techniques with sentiment analysis of user-generated content from social media. By analysing user sentiments and preferences, the algorithm refines recommendations to match users' dynamic interests more effectively. The performance of CoFiSent is evaluated against existing algorithms, including standard collaborative filtering and content-based filtering models. Simulations are conducted using real-world datasets, and the system's effectiveness is measured through precision, recall, F1-score, and mean reciprocal rank (MRR). Results demonstrate that CoFiSent outperforms traditional methods, particularly in capturing emerging trends and contextual preferences reflected in social media activities. The integration of multiple data sources enhances the overall accuracy and relevance of recommendations, offering a significant improvement in user satisfaction. The research findings suggest that incorporating social media sentiment and search behaviour into film recommendation systems represents a promising approach for next-generation entertainment platforms.