The intellectual system of movies recommendations based on the collaborative filtering

Abstract

The investigation deals with designing and developing of intellectual system of movies recommendations  based on the collaborative filtering using the Python software environment. In particular, the approaches (Content-based approach, Collaborative filtering, Hybrid models) in recommendatory system construction with the help of neural networks have been analyzed. It has been established that it is difficult to implement and learn the Content-based approach and it strongly depends on the subject area. Collaborative filtering is more simple in implementation, training, it is universal, but it has a flaw in the form of a «cold-start». Accordingly, the collaborative filtering has been chosen for the design and development of the intellectual system of movies recommendations. While designing a system of recommendations based on collaborative filtering, the Naive Recommendations, Recommendations based on average ratings of similar users, Recommendations based on average user ratings and similarity matrix have been described; their algorithm and their implementation using the Python software environment have been demonstrated. As a result the intellectual system of recommendations has been realized and it can offer a movie to the user according to his/her preferences.