The goal of this study is to predict urban growth in Cox’s Bazar, Bangladesh, through 2050 using historical land-use patterns, geospatial data, and machine learning techniques. Multi-temporal Landsat imagery (1990–2020) was used to classify land cover into urban, vegetation, water, and bare land. Key drivers influencing urban expansion, population density, distance to roads, elevation, slope, distance to coast, and proximity to existing urban areas were extracted and processed in a geospatial environment. By analyzing these factors together, this study identifies patterns and trends in urban expansion that can inform planning decisions. The approach also highlights areas at risk of rapid, unplanned growth, providing a basis for proactive urban management and sustainable development strategies. Random Forest (RF) will be used to quantify the relative importance of candidate drivers of urban growth, and a Cellular Automata–Markov framework will be applied to simulate future urban expansion using transition probabilities derived from multi-temporal land-use/land-cover (LULC) maps. Model validation will be conducted using Kappa statistics and strong overall accuracy. The simulations will generate projected urban extent for multiple future time horizons (2030, 2035, and 2040). Hotspot analyses will be used to detect areas of rapid urbanization, which are hypothesized to cluster along major road corridors, at peri-urban interfaces, and around refugee settlements. This study demonstrates that combining GIS, remote sensing, and machine learning provides quantitative, data-driven forecasts that can guide sustainable urban planning, infrastructure development, and environmental management in rapidly growing coastal cities.