15. Landslide susceptibility mapping by using frequency ratio method and machine learning models: A case study of Phin Ngan Commune, Lao Cai Province, Vietnam

Abstract

Landslide susceptibility assessment in Phin Ngan commune, Lao Cai province, Vietnam has been carried out by applying three machine learning (ML) models (Logistic Regression - LR, Bayesian Network - BN and Support Vector Machines - SVM), and the Frequency Ratio (FR) method. First, FR was applied to calculate the weighting values of each parameter classes in the factor maps. Second, ML models were applied to calculate the importance of each landslide related factor map. Then, the landslide susceptibility index (LSI) maps were generated by combining the importance values of factor maps obtained from ML models and the parameter classes that was assigned the weighting values of factor maps created by FR model. Next, the performance of these ML methods has been evaluated using receiver operating characteristic (ROC) curve and area under the curve (AUC). Analysis and comparison of the results showed that all three ML models performed well for landslide susceptibility assessment (AUC = 87.2 % - 97.5 %). The BN model (AUC = 97.5 %) showed the best performance in comparison to other landslide models (LR model: AUC = 94.6 % and the SVM model: AUC = 87.2 %). The results indicated that the models have given outputs with good forecasting ability. They are also very useful in supporting land - use planning, the prevention and mitigation of risks due to landslides in the research area.