Machine Learning-Driven Multimodal Feature Extraction and Optimization Strategies for High-Speed Railway Station Area

Abstract

The construction of high-speed railway (HSR) station areas serves as a crucial catalyst for urban spatial evolution. However, the absence of targeted urban management theories has led to widespread spatial resource waste and post-construction abandonment phenomena in these areas. Existing research predominantly focuses on development strategies for individual construction elements of HSR stations yet lacks comprehensive strategy formulation through coordinated multi-level elements from a sustainable perspective. This study establishes a national database comprising 1018 HSR station area samples across China in 2020, integrating built environment characteristics, HSR network topology, ecological considerations, and socioeconomic indicators. Guided by the land equilibrium utilization theory, we employ the random forest Boruta algorithm to identify critical features, using land supply capacity and development intensity as target variables. Subsequently, K-means++ clustering analysis based on these key variables categorizes the samples into nine distinct clusters. Through normal distribution tests, we establish reference ranges for cluster-specific indicators and propose tailored development strategies across multiple dimensions. This research develops a multimodal feature extraction and evaluation framework specifically designed for the large-scale analysis of HSR station areas. The nine-category strategic recommendations with defined quantitative threshold intervals provide decision-makers with visually intuitive, operationally implementable, and practically significant guidance for spatial planning and resource allocation.