Developing artificial intelligence algorithms for urban functional analysis and planning

This study aims to develop a new model for the classification and prediction of urban functions by integrating artificial intelligence and spatial data science techniques within the framework of urban geography. In the functional classification analysis conducted for İstanbul, nearly one million open-source spatial datasets compiled from Points of Interest (POI) and building footprints were utilized. Mixed-use parcels were identified using the Shannon Entropy technique, while the spatial distribution and detection of dominant functions were determined through the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Subsequently, spatial variables were generated from the environmental characteristics of parcels with known functions. These variables were employed in the Başakşehir and Sancaktepe districts to predict the possible functions of vacant parcels using the Extreme Gradient Boosting (XGBoost) machine learning algorithm. The functional classification accuracy analysis for İstanbul yielded an overall accuracy of 94.8% and a Kappa coefficient of 0.94, indicating that the classification effectively represented the functional spatial patterns. In the selected districts of Istanbul, the XGBoost model achieved an overall accuracy of 93%, with macro average precision, recall, and F1 scores ranging between 0.90–0.91, demonstrating balanced predictive performance across different function types. The results indicate that functional diversity is concentrated in the central business districts, coastal zones, and district centers of İstanbul, reflecting the city's mixed-use and multifunctional character in these areas. As the distance from these central and coastal zones increases, land use becomes more homogeneous, dominated mainly by residential function. The housing function dominates citywide, followed by commercial, public service, and industrial uses. Prediction analyses indicate that a substantial proportion of vacant parcels tend to transform into residential, commercial, and recreational functions. The developed AI-based classification and prediction model provides a data-driven, rapid, and objective tool for urban planning processes. Accordingly, this study establishes a scientific foundation for the sustainable, balanced, and accessible development of cities.