MOVIE RECOMMENDATION USING MODIFIED NEURAL COLLABORATIVE FILTERING AND FP-GROWTH ALGORITHM

Recommendation system is a process of suggesting more likely items to the users based on their preferences and interest. Applications of recommendation system are seen in almost many areas like e- commerce, social media and multimedia platform. In recent days, the hybrid collaborative filtering techniques are used for the recommendation to improve the suggestions for users. Nowadays, collaborative filtering using neural networks is used in recommendation process. This paper aims to develop a modified neural collaborative filtering model to recommend movies to users. Movies are rated by the users in the scale of 1 to 5. Users prefer to watch movies not only based on the ratings but also like other factors like genre, cast, crew, etc. In this paper genre is also considered along with the ratings to train the neural model. The user-movie interaction matrix used by many collaborative filtering techniques suffers from sparsity. To overcome this problem, the modified neural collaborative filtering model is used to find the ratings of the movies not watched by the user. Finally a recommendation module using FP-Growth algorithm is developed to suggest users with movies considering ratings for the movie and genre of the movie. The loss function, mean absolute error is used to analyze the performance of the modified neural collaborative filtering model. The testing loss of trained model is found to be 0.1017. The recommendation module is evaluated using Normalized Discounted Cumulative Gain (NDCG) metric. The experimental results show that the proposed recommendation system performs better compared to Nearest neighbor and correlation based recommendation systems.