Applying different resampling strategies in machine learning models to predict head-cut gully erosion susceptibility

Gully erosion is one of the advanced forms water erosion. Identifying effective factors and gully predicting important tools to control manage such phenomenon. The main purpose this study evaluate effect four different resampling algorithms including cross-validation (5-fold 10-fold) bootstrapping (Bootstrap Optimism bootstrap) on boosted regression tree (BRT), support vector machine (SVM), random forest (RF) models in spatial modeling evaluation head-cut Konduran watershed. For purpose, based an extensive field survey, points were identified first, a map distribution area was prepared. Then 18 variable identify prepare as affecting occurrence To assess efficiency models, receiver operating characteristics (ROC) under curve (AUC) used. The results indicated that use increases models. integrated optimism-bootstrap-BRT, optimism-bootstrap-SVM, Optimism-Bootstrap-RF with AUC 0.85, 0.823 0.89 respectively, outperformed 5fold (BRT, SVM, RF), Cross-validation 10fold RF) Bootstrap algorithms.