Towards the Implementation of Movie Recommender System by Using Unsupervised Machine Learning Schemes

Abstract

This study aimed at finding out the similarity to create a movie recommendation system and grouping based on the user. The purpose of the recommendation system as information for customers in selecting films according to features. The recommendation system can be performed with several algorithms as a grouping such as K-Means, K-Means Mini Batch, Birch Algorithm, Affinity Propagation Algorithm and Mean Shift Algorithm. We recommend methods to optimize K as a precaution in increasing variance. We use clustering based on Movie ratings, Tags, and Genre. This study would find a better method and way to evaluate the clustering algorithm. To check the recommendation system, we utilize social network analysis and mean squared error to explore the relationships between clusters. We also utilize average similarity, computation time, and clustering performance evaluation in getting an evaluation as a comparison of the recommendation system. Clustering Performance Evaluation with Silhouette Coefficient, Calinski-Harabasz, Davies-Bouldin.