<![CDATA[The survival of living things is highly dependent on the important role of air. Clean air that is free from pollution is a standard for a quality environment that supports life. The Machine Learning approach can be an alternative in conducting data-based air pollution monitoring to assist in making the right decisions to deal with air pollution early on. This research aims to optimize the performance of the Light Gradient Boosting Machine (LightGBM) algorithm in air pollution classification combined with PSO optimization. The LightGBM or Light Gradient Boosting Machine algorithm is a Gradient Boosting algorithm that has decision tree-based learning, but in its application, LightGBM is prone to overfitting because it is sensitive to hyperparameters. Therefore, optimization techniques are needed to maximize performance. Particle Swarm Optimization (PSO) is an optimization method inspired by the movement of flocks of birds searching for optimal solutions. The data used is the Air Pollution Standard Index data. The research method includes data collection, data preprocessing, splitting the data, PSO optimization, model training, and model evaluation. The results show that PSO optimization can improve the performance of the LightGBM model. The LightGBM model with PSO optimization produced an evaluation matrix with an accuracy of 0.9510, precision of 0.9256, recall of 0.9261, and F1-score of 0.9247, demonstrating the model's ability to accurately classify air pollution. Meanwhile, the LightGBM model without optimization produced an evaluation matrix with an accuracy of 0.9455, precision of 0.9201, recall of 0.9170, and F1-score of 0.9182.]]>
Copyrights © 2025
