Predicting Urban Expansion using Cellular Automata and Machine Learning: A Multi-Model Evaluation Framework

Abstract

Rapid urbanization has led to a significant land use land cover (LULC) evolution necessitating the robust predictive framework to guide sustainable urban planning. This study presents a comparative analysis of urban growth modeling using a hybrid cellular automata (CA) approach integrated with multiple machine learning techniques including random forest (RF), support vector machine (SVM) and Extreme Gradient Boosting (XGBoost). A tier-1 Indian city was selected as the study area (Jaipur city). Ancillary factors such as proximity to roads, rivers, elevation, slope and population density were incorporated to enhance the transition potential modeling. Model performance was evaluated using metrics such as Kappa coefficient, overall accuracy and F1 score. The results indicated that the CA-RF model outperformed the others in terms of both spatial realism and statistical accuracy, followed closed by CA-XGBoost. Spatial comparison of predicted outputs also revealed variations in growth direction, pattern compactness and built-up density. This research underscores the significance of combining CA with data-driven ML approaches for capturing complex urban dynamics and provides a decision-support framework for urban planners and policymakers.