Machine Learning Models for Land Cover Change Detection: A Sentinel-2 and Google Earth Engine Study

This study develops and identifies the optimal machine learning model for land use and land cover (LULC) classification using Sentinel-2 imagery and Google Earth Engine (GEE) between 2018 and 2023. Among the Classification and Regression Trees (CART), Random Forest (RF), and Gradient Boosting (GB) models, the Random Forest (RF) model demonstrated superior performance. In 2023, it achieved an overall accuracy of 97.89<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> and a Kappa coefficient of 0.9699. After optimization through hyperparameter tuning, the Random Forest (RF) model improved to 97.92% accuracy and a Kappa coefficient of 0.9704. The resulting LULC maps revealed significant changes in residential, industrial, and vegetation areas, highlighting the effectiveness of machine learning for sustainable land management and planning.