Film Recommendation System using Neural Collaborative Filtering Method

Abstract

Films are growing rapidly thanks to the increasing number of internet users. According to USC (University of Southern California), users easily get bored if they cannot find the movie they want within 1.5 minutes or within 90 seconds and may switch to other platforms, resulting in a decrease in active users. This problem can be overcome by using a recommendation system that can provide film recommendations based on user ratings. In this design, the recommendation system used is Neural Collaborative Filtering (NCF). Testing using 3,235 film titles, 610 users, and 87,812 ratings produced the best scenario with a recall value of 69.6% and Normalized Discounted Cumulative Gain (NDCG) with a value of 81.4%. From the test results, it can be concluded that 69.6% of films that should have been recommended were successfully recommended, and the recommendation position was similar to the actual position in 81.4% of cases.