Comparison of user-based and item-based collaborative filtering methods in recommender system

Nowadays, movies have become one of people's favorite entertainment. The number of films that reach thousands or even millions make it difficult for film fans to choose the right film to watch as desired. A recommendation system is required to provide recommendations on which movies to watch as they want. A recommendation system is a system that assists users in dealing with enormous amounts of information by giving particular recommendations for users, with the goal that these recommendations would fulfill the users' wishes and needs. Based on filtering, categorizing items, and information that takes preferences from user behavior and history, the recommendation system can discover and present content with a high potential for people to choose or utilize. Recommendation systems are classified into three types based on the approach used: collaborative filtering, content-based filtering, and hybrid filtering. Collaborative filtering, one of the approaches commonly employed in recommendation systems, was applied in this study. There are two ways in the Collaborative Filtering approach: User-Based Collaborative Filtering (UBCF) and Item-Based Collaborative Filtering (IBCF). This study aims to determine which system produces the most significant results between UBCF and IBCF by comparing predicted ratings to actual ratings using 80 percent training data and 20% testing data. The data set taken from MovieLens.org consists of 993920 ratings provided to films by users. The MovieLens data collection will be analyzed with the R program and the R package recommender lab. The results reveal that the IBCF model outperforms the UBCF model in prediction error, with RMSE, MSE, and MAE values of 1.133177, 1.284091, and 0.852272, respectively. As a result, the UBCF Model is the best recommendation model for predicting each user's rating