Movie Recommendation System Based on SVD Collaborative Filtering

With the development of the film industry, more and more people enjoy watching movies and regard them as an indispensable entertainment activity in their life. Nowadays, there is a tremendous amount of information about movies on the Internet, and people can get lost in it if they don't choose from it. However, manual selection and browsing movie information can also be time-consuming and laborious. To solve this problem, people need a movie recommendation system. The movie recommendation system can recommend movies that users may like by calculating the similarity between users and films. This paper applied the collaborative filtering algorithm, which is widely used nowadays. The main problems that recommendation systems now face are cold start and sparsity. This research adopted Singular Value Decomposition (SVD) based collaborative filtering system to solve the issues. This project first predicted the performance of two classic collaborative filtering systems (User-based collaborative filtering and Item-based collaborative filtering) based on the MovieLens ML-100K dataset. Then RMSE and MAE were used as evaluation indexes to evaluate the effects of the three recommendation algorithms. This research found that SVD had the best recommendation effect under this dataset through experiments. Finally, by changing the k value, this paper found that within a specific range, the recommendation effect of SVD is improved with the increase of the k value.