Thumbnail Personalization in Movie Recommender System

Personalization of the user experience is a key aspect of increasing user engagement and retention on online platforms. In the proposed work, a hybrid recommender system combines content-based filtering using cosine similarity and collaborative filtering using triangle similarity. In each system, a predicted score is calculated for a given user and film. These approaches are combined by taking a weighted average of their individual results to improve accuracy. These error values match and outperform those of other existing systems. The system assigns a specific thumbnail to the movie recommended based on preferences of the user to the actors featured on the artwork. A single artwork is selected among many through a weighted probability. The aim of the proposed work is to build a system whose personalization techniques can be accessible to smaller scale platforms and can be built upon to further enhance user experience on user-centric platforms.