The improved mountain gazelle optimizer for spatiotemporal support vector regression: a novel method for railway subgrade settlement prediction integrating multi-source information

The uneven settlement of railway subgrades not only affects the comfort of train operations but, in extreme cases, may compromise operational safety. As a result, accurately predicting subgrade settlement is crucial for maintaining both safety and operational efficiency. This study introduces an Improved Mountain Gazelle Optimizer for the Spatiotemporal Support Vector Regression (IMGO-STSVR) model, which effectively predicts railway subgrade settlement. Data are collected using Permanent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) technology in combination with a multi-source environmental monitoring system. The proposed improvement to the Mountain Gazelle Optimizer (IMGO) enhances the model’s optimization capabilities, while the Support Vector Regression model is improved by the constructed spatiotemporal kernel function (STSVR). Experimental results demonstrate that the IMGO-STSVR model achieves high accuracy and stability across various experimental sites. This method provides valuable insights for predicting subgrade settlement in the railway industry, aiding in the early identification of potential risks, optimizing maintenance strategies, and ensuring the safe and efficient operation of rail transport.