Web Based Video Games Recommendation System Using Collaborative Filtering Method

The video games market is enormous and continues to grow, both in terms of quantity and variety. In this way, an effective and efficient recommendation system is required to provide the right choices for the user. This research aims to conduct a comparison of collaborative filtering recommendation system based on matrix decomposition. These techniques analyze the collective preferences and behaviors of users, thereby enhancing the accuracy and personalization of game recommendations. The matrix factorization algorithms being compared are Alternating Least Squares (ALS), Singular Values Decomposition (SVD), and Non-Negative Matrix Factorization (NMF), each of which has its own characteristics and approaches for extracting hidden features from the data. The evaluation results show that the NMF model obtained the best results for the evaluation metrics Root Mean Squared Error (RMSE) with a value of 1.34, Mean Absolute Error (MAE) with a value of 0.68, precision@k with a value of 0.90, recall@k with a value of 0.56, and F1- score@k with a value of 0.69. Meanwhile, from the duration of modeling process, the fastest results were obtained by SVD with a time of 0.15s. The best model, NMF, was then used as the basis for developing a video game recommendation system website. Website performance is then tested using the Black Box Testing method and gets good results for every aspect of the test.