Modeling the Spatial Suitability of Urban Parks Using the Random Forest Regression Algorithm: A Case Study of Fardis City, Iran

Objective: Rapid population growth and limited urban green spaces in densely populated cities like Fardis highlight the need for data-driven approaches to optimize urban park planning. This study aims to predict spatial suitability for urban parks using environmental and spatial factors. Methods: Fifteen spatial and environmental variables, including vegetation index, population density, water resource accessibility, slope, elevation, erosion, land use, and other relevant environmental factors, were used. All data correspond to the year 2025 and were derived from Sentinel-2 and Planet satellite imagery as well as geospatial data of Fardis city. Data were normalized using the Min–Max method and split into training (80%) and testing (20%) datasets. A Random Forest model was trained with 200 trees, a maximum depth of 20, minimum 3 samples per leaf, and the number of features considered for each split set to the square root of the total features. Results: Model evaluation demonstrated high predictive accuracy with a coefficient of determination (R² = 0.968), root mean squared error (RMSE = 0.090), and mean absolute error (MAE = 0.040). Spatial analysis of the results indicated that urban park development should be prioritized in the northeast and east of Fardis, while supportive actions are recommended in the southwest to enhance green space quality and accessibility. Conclusion: The proposed framework provides an efficient tool for data-driven decision-making, reducing spatial inequalities and promoting sustainable urban development.