Tags:ADABoost, Bagging, Boosting, J48 Decision Tree, Landslide Susceptibility, Random Forest, Rotation Forest and XGBoost
Abstract:
mplementing the various machine learning algorithms for landslide susceptibility mapping has been researched by many authors and is worth considering the issue. In the present study, the effectiveness of Decision Tree and its bagging and boosting based ensemble model techniques (like Random Forest, Rotation Forest, Extra Tree, Adaboost, and XGBoost) has been evaluated via generating the Landslide Susceptibility Map (LSM). Both threshold based i.e. overall accuracy and rank based i.e. Area Under Receiver Operating Characteristics(AUROC) measures have been used as the criteria for evaluating the various model’s performance. The result concluded that the XGBoost model has outperformed the other implemented algorithms after performing hyper-parameters tuning for each algorithm. The study area considered for the present study is Rishikesh to Gangotri axis with a buffer area of 3 km on each side. It is the first time that these algorithms have been implemented and compared for this study area.
Landslide Susceptibility Mapping Using J48 Decision Tree and Its Ensemble Methods for Rishikesh to Gangotri Axis