A Comparison Study of Model Based Collaborative Filtering Using Alternating Least Square and Singular Value Decomposition
Hartatik Hartatik, Bayu Permana Sejati, Aulia Nur Fitrianto, Wiwi Widayani · 2021
Recommender systems are systems that are built to recommend items to users based on a variety of criteria. These systems predict the most likely product that users are likely to buy and are interested in it. The most used recommender system is collaborative filtering. In this research, the author proposes determining the quality of the recommender system by comparing model-based collaborative filtering techniques, i.e., Alternating Least Squares and Singular Value Decomposition with three different characteristics dataset. This study showed Model-Based Collaborative Filtering (Alternating Least Squares and Singular Value Decomposition) able to improve the quality of the recommender system by considering the hyperparameter tuning result. Alternating Least Squares performs slightly better than Singular Value Decomposition in MovieLens and jester dataset.