Comparative Assessment of Machine Learning Algorithms in Characterizing Urban Sprawl in Benin City

Urban growth across Sub-Saharan Africa is occurring at an unprecedented pace, often without adequate planning control. In cities such as Benin City, Nigeria, this expansion is reshaping landscapes, accelerating forest loss, and increasing environmental vulnerability. This study comparatively assesses two widely used machine learning algorithms: Support Vector Machine (SVM) and Random Forest (RF), in mapping and characterizing urban sprawl using multi-temporal Landsat imagery from 2014, 2019, and 2024.Supervised classifications were performed to generate land use/land cover (LULC) maps, and classification accuracy was evaluated using confusion matrices, Overall Accuracy (OA), Kappa statistics, and class-level accuracies. Post-classification change detection, land cover transition analysis, transportation corridor buffering, and Digital Elevation Model (DEM) assessment were employed to examine spatial growth patterns and their environmental implications. The results show that built up land increased dramatically from approximately 13% to 18% in 2014 to about 35% to 36% in 2024. Conversely, forest cover declined from over 47% to 56% to less than 12% to 16% during the same period. Nearly 79% of new urban development occurred within 1 km of major road corridors, while about 68% took place in medium elevation zones. Although both algorithms performed strongly, SVM demonstrated more consistent and stable classification accuracy across years. The findings highlight that the scale of urban transformation in Benin City and underscore the value of machine learning based monitoring in supporting sustainable urban planning and environmental management.