Optimizing Alternating Least Squares for Recommender Systems Using Particle Swarm Optimization

Recommender systems play a crucial role in various digital platforms by assisting users in discovering relevant items. The research problem addressed in this study is the limited predictive accuracy of ALS-based recommender systems due to suboptimal parameter selection. This study explores how Particle Swarm Optimization (PSO) can be leveraged for parameter optimization to address this limitation. The dataset used is MovieLens 1M, which contains over one million user ratings for thousands of movies. The research process includes data preprocessing, data splitting, model training, and evaluation using Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) as the primary metrics. The evaluation results indicate a significant improvement in model performance after optimization, with RMSE decreasing from 0.895 to 0.860 and MAE from 0.704 to 0.680. These findings demonstrate that optimization algorithms can effectively improve the prediction accuracy of recommendation systems. This research contributes to the application of swarm-based optimization techniques in enhancing matrix factorization-based recommender systems.