A new approach to combating landslides: Optimum weighted ensemble learning (OPWE)

Abstract

This study is critically important in the early prediction of landslides, aiming to minimize the impacts of natural disasters. The development of early warning systems is vital to prevent loss of life and property. This study aims to predict landslide occurrences by using an ensemble learning model to predict landslide probability. Furthermore, a new ensemble learning approach called OPAT, based on Hybrid K-Means- Battle Royale Optimization Algorithm (HBROA) is proposed to improve the accuracy of the predictions. The weighted voting method is used instead of Majority Voting, which is commonly used in ensemble learning studies, and the classifier weights used in this method are optimized with the help of a hybrid metaheuristic algorithm called HBROA. In the study, basic classifiers such as Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) are used and the results of these classifiers are combined with the bagging method. The proposed method outperformed the baseline classifiers and other methods in the existing literature with an accuracy of 81.89%. As a result, this study presents a new approach to landslide prediction using different classifier methods. Furthermore, an ensemble learning approach based on metaheuristic algorithms is proposed to identify landslide-prone areas and mitigate potential damages more accurately. This new method represents a significant step forward in detecting landslides in advance and minimizing the impacts of such natural disasters.