Comparison of Multi-Criteria Decision Making, Statistics, and Machine Learning Models for Landslide Susceptibility Mapping in Van Yen District, Yen Bai Province, Vietnam

Abstract

Landslides are natural hazards that pose a significant threat to human lives and infrastructure. Landslide susceptibility mapping aims to classify areas at risk of landslides. Multi-Criteria Decision Making (MCDM) algorithms have the advantage of incorporating expert opinions, while Statistics and Machine Learning models demonstrate greater objectivity. This study compares three representative models, namely Analytic Hierarchy Process (AHP), Frequency Ratio (FR), and Random Forest (RF), for developing a landslide susceptibility model in Van Yen District, Yen Bai Province. The classification points for landslides were divided into a 70% training set and a 30% testing set. Thirteen conditioning factors were used to evaluate the landslide's influences. The results show that the AHP and FR models perform well with AUC = 0.842 and AUC = 0.852, respectively, while the RF model outperforms them with AUC = 0.949. The study demonstrates the applicability of these models for analyzing landslide susceptibility in the research area, highlighting the strong potential of machine learning models.