Exploration of bipartite graph to improve recommendation task

Abstract

By using bipartite graphs in movie recommender systems, we can address the challenge of recommender system limitations (sparsity, scalability, cold start, etc.) to improve recommendation personalization. Traditionally, recommendation systems use methods such as collaborative filtering, content-based filtering, hybrid filtering and others to understand user preferences and suggest similar movies based on past ratings. However, when a new user joins the system or when a new movie is added to the database, a problem arises. Bipartite graphs offer a promising solution to this problem. By representing users and movies as separate nodes in a bipartite graph, we can use similarity calculation algorithms (cosine similarity and others) to recommend movies to new users based on the preferences of similar users. For example, by looking at past interactions between users and movies, we can identify communities of users who share similar tastes. We can then recommend movies to new users based on the preferences of similar users within those communities. By integrating contextual information such as user ratings, movie genres, actors, directors, and release years, we can improve the relevance and personalization of movie recommendations. Using bipartite graphs, we can also efficiently manage new movies by linking them to users with similar preferences, even without direct ranking. In conclusion, the use of bipartite graphs in movie recommend systems presents an innovative approach to overcoming the limitations of traditional recommend systems, improve the personalization of recommendations, and provide relevant suggestions even for new users and new movies.