EVALUATION OF THE TOPOGRAPHIC CHARACTERISTICS AT THE LANDSLIDE OCCURRENCES BASED ON COVARIANCE STRUCTURE ANALYSIS AND MACHINE LEARNING

Abstract

 This study evaluated the topographic characteristics at the landslide outbreak locations based on the covariance structure analysis and the machine learnings. The support vector machine (SVM) and random forest (RF) were employed in this study to achieve the objective. The estimated results were compared to investigate the accuracy of the detection of landslide occurrences. The SVM and RF gave reasonable results if a technique for oversampling is applied to correct for a bias in the training dataset. Lastly, the discriminated results were projected onto a map to evaluate the possibility of landslide occurrences.