Movie Recommender System Using Cascade Hybrid Filtering with Convolutional Neural Network

Abstract

The current technological advancements have made it easier to watch movies, especially through online streaming platforms such as Netflix. Social media platforms like Twitter are used to discuss, share information, and recommend movies to other users through tweets. The user tweets from Twitter are utilized as a film review dataset. Film ratings can be used to build a recommendation system, incorporating Collaborative Filtering (CF) and Content-based Filtering (CBF). However, both methods have their limitations. Therefore, a hybrid filtering approach is required to overcome this problem. The filtering approach involves CF and CBF processes to improve the accuracy of film recommendations. No current research employs the Cascade Hybrid Filtering method, particularly within the context of movie recommendation systems. This study addresses this gap by implementing the Cascade Hybrid Filtering method, utilizing the Convolutional Neural Network (CNN) as the evaluative instrument. This research presents a significant contribution by implementing the Cascade Hybrid Filtering method based on CNN. This research uses several scenarios to compare methods to produce the most accurate model. This study's findings demonstrate that the application of Cascade Hybrid Filtering, incorporating CNN and optimized with RMSProp, yields a movie recommendation system with notable performance metrics, including an MAE of 0.8643, RMSE of 0.6325, and the highest accuracy rate recorded at 86.95%. The RMSprop optimizer, facilitating a learning rate of 6.250551925273976e-06, enhances accuracy to 88.40%, showcasing a remarkable improvement of 6.00% from the baseline. These outcomes underscore the significant contribution of the paper in enhancing the precision and effectiveness of movie recommendation systems.