Unlocking patterns in urban land use efficiency: A global analysis using XGBoost and Bayesian networks

Land use efficiency has emerged as a critical consideration for urban sustainability, but its underlying drivers remain insufficiently explored at the global scale. This paper addresses this gap by combining XGBoost and Bayesian Networks to identify and model the principal factors influencing urban land use efficiency worldwide. XGBoost isolates key drivers and quantifies their effects, Bayesian Networks provide a probabilistic framework for capturing complex dependencies among these variables. The analysis highlights clear threshold effects, showing that modest positive population growth yields substantial efficiency gains, while both stagnation and rapid growth can undermine outcomes. Similarly, systematic redevelopment and expansion of built-up areas consistently increase efficiency, particularly when aligned with demographic dynamics. By shedding light on how different configurations of growth and development translate into higher or lower urban land use efficiency, the study offers actionable guidance for decision makers. • XGBoost and Bayesian Networks reveal complex patterns in global urban land use. • Moderate population growth and recent urbanisation lead to efficient land use. • Balanced strategies should manage growth and encourage compact urban design.