A Machine Learning-Driven Decision Support System for Urban Expansion Prediction Using Multi-Temporal Remote Sensing Data

Abstract

With the acceleration of urbanization, resource consumption and environmental pressure caused by urban expansion have become increasingly prominent, highlighting the urgent need for a scientific planning decision support system. This study proposes a machine learning-driven decision support system for urban expansion prediction using multi-temporal remote sensing data, aiming to enhance the scientificity and effectiveness of urban planning. Key technologies include image enhancement, noise reduction, and feature extraction to obtain high-quality urban expansion data, with specific algorithms such as Convolutional Neural Networks (CNN) for spatial feature learning and Iterative Slow Feature Analysis (ISFA) for change detection. The analysis found that the land use efficiency in some areas has been significantly improved, and the related change range is 11.1% (representing an 11.1% increase in land use efficiency compared to the baseline). In the correlation analysis between economic development and traffic flow, the change range reached 39.4 (an index value indicating the strength of correlation, with higher values representing stronger associations). The average change rate of urban areas is 56.3%, while green space coverage changes by 73.2% and building density increases by 68%. The average annual change rate of urban areas is 56.3% over the period 2010–2020, while the effective change detection rate reaches 92%.