Hybrid Recommender System Using Artificial Bee Colony Based on Graph Database

Recommender systems hoard appropriate suggestions guided by the interactions and previous choices of the users. In a world of the ever-growing amount of data, we are overpopulated with undesired information, which is making it difficult to operate or choose. Hence, we require a recommender system that is enough capable to deduce desired suggestions which can be valuable for the users. Thus, they are rising in popularity and becoming part and parcel of day-to-day activities in our general life. Here, in this paper, we implement the recommender system, which is conceptualized on a hybrid filtering algorithm that helps in dealing with limitations of both content and collaborative filtering reinforced with artificial bee colony optimization along with k- nearest neighbor for better performance. For this purpose, we used MovieLens dataset, which contains information regarding users, movies, and ratings given by the users. Here, we gathered a pre-filled user project scoring matrix and have compared multiple models of recommender systems for their precision and recall factor. We are using a recommender system based on graph database which uses graph traversal and pathfinding algorithms to establish relations and hence is more robust and faster in implementation. The thickness of the edges connecting movie nodes and the indegree of a movie node specifies the recommendation limit of the movie. The experiment results on MovieLens dataset establishes scope for future scalable models and delivers competent outcomes brought in comparison with traditional systems. Preliminary results show improvement of 9% in precision and 3% improvement in recall over traditional systems.