In achieving sustainable urban development, the intricate dynamics of spatio-temporal land use changes and their regional driving forces necessitate comprehensive investigation, particularly in developing countries. This study aims to analyze the detection of land use and land cover (LULC) changes, focusing on urban sprawl and future predictions. The performance of five supervised machine learning algorithms for LULC classification in Larache City, Morocco, during 2015–2021 is evaluated, addressing the research question of which algorithm best captures LULC changes accurately. Based on the theoretical foundation that accurate LULC mapping informs sustainable resource management, a theoretical framework rooted in remote sensing and machine learning principles is employed. The research methodology involves the analysis of archived Sentinel 2 imagery to detect LULC changes, incorporating change detection modelling and metrics such as overall accuracy and Kappa coefficient. Results highlight the superiority of the Support Vector Machine (SVM) algorithm, with an average overall accuracy of 93.67% and a Kappa coefficient of 0.93 across the study years. In contrast, the Classification and Regression Tree (CART) algorithm achieves a lower accuracy of 82.67% and a Kappa coefficient of 0.79. The implications are substantial. Accurate LULC classification is pivotal for effective urban planning and resource management, especially in developing countries. The accuracy of SVM underscores its potential as a robust tool for developing LULC maps, aiding decision-makers in land management strategies. This study contributes to understanding LULC dynamics, urban sprawl, and future projections, thereby providing essential data for informed urban development decisions and sustainable land use strategies. As cities in the developing world evolve, integrating precise LULC insights becomes paramount for achieving balanced urban growth and environmental conservation.