Movie Recommendation System Model using Bisecting K-Means Technique and Collaborative Filtering

In the current film industry, the competition is very big. We can see it in online streaming content through the ratings obtained. Film itself is a visual work that is packaged as a product of public entertainment for a specific purpose. However, there are also many films that are considered not to meet the audience's expectations. Even the films presented are sometimes illegal or pirated films. We can also find out whether a film is recommended or not. The problem is that viewers rarely understand how to see recommendations or even provide appropriate film recommendations. This study aims to develop a film recommendation system model using a combination of K-Means bisecting and Collaborative Filtering. The film data used in this study comes from Movie-Lens which consists of 100,000 ratings from 668 users for 10329 film titles in 18 film genres. The training process consists of a cluster process with the K-Means bisecting algorithm and calculating similarity values ​​with collaborative filtering (item-based and user-based). The testing process is carried out to calculate the system error value by calculating the Mean Absolute Error (MAE) value. The results of the study show that recommendations with bisecting K-Means and user-based collaborative filtering get lower MAE values ​​compared to bisecting K-Means and item-based collaborative filtering.