Evaluation modeling of highway collapse hazard based on rough set and support vector machine.

Area of Science:

• Geotechnical Engineering
• Environmental Science
• Data Science

Background:

• Landslide prediction is critical for mountain highway construction safety.
• Existing models face challenges with small datasets and complex influencing factors.

Purpose of the Study:

• To develop a robust landslide hazard prediction model for mountainous highway construction.
• To enhance prediction accuracy and computational efficiency using data-driven methods.

Main Methods:

• A hybrid model integrating Rough Set Theory (RST) and Support Vector Machine (SVM) was developed.
• RST was used to reduce landslide influencing factors to nine key indicators.
• SVM was trained and optimized using field data from post-earthquake landslides.

Main Results:

• The RS-SVM model achieved 100% accuracy in cross-validation and testing.
• Key factors influencing collapse activity were identified: slope shape, aspect, gradient, height, structural face exposure, lithology, weakness/free face relationship, vegetation cover, and rock weathering.
• The model demonstrated superior performance in handling small and nonlinear datasets.

Conclusions:

• The RS-SVM model offers an accurate and efficient solution for landslide hazard assessment in mountainous regions.
• This approach provides a valuable framework for future landslide risk management and mitigation strategies.