Optimizing the Sample Selection of Machine Learning Models for Landslide Susceptibility Prediction Using Information Value Models in the Dabie Mountain Area of Anhui, China

The evaluation of landslide susceptibility is great significance in the prevention and management geological hazards. accuracy prediction model based on machine learning significantly higher than that traditional expert knowledge conventional mathematical statistics model. correct reasonable selection non-landslide samples greatly improves reliability regional Focusing problem selecting for evaluation, this paper proposes a method combination an information statistics. First, influence factors are screened by correlation analysis method. Second, value used to delimit areas with low relatively susceptibility, points randomly selected. Third, combined IV-ML, such as logistic regression (IV-LR), random forest (IV-RF), support vector (IV-SVM), artificial neural network (IV-ANN), established. Finally, Dabie Mountain area Anhui Province analyzed, results using IV-LR, IV-RF, IV-SVM, IV-ANN LR, RF, SVM, ANN methods compared. evaluated examining ACC, AUC, kappa values indicate effect IV-ML models (IV-LR, IV-ANN) ML (LR, ANN).