Research on the application of Breiman's algorithm integrated with the Random Forest in determining the importance of input factors to landslide formation in Son La province

Abstract

Landslide susceptibility maps are effective and intuitive tools in natural disaster management, helping to minimize damage caused by natural disasters due to the specific spatial information they provide. However, the performance of these susceptibility maps depends mainly on the number and importance of input factors. Determining the importance and order of influencing factors often receives little attention in landslide prediction studies. Breiman's algorithm, integrated into the Random Forest method, can comprehensively determine the importance and order of input variables by considering the correlation relationship between the landslide inventory map and these input variables. Consequently, this study utilized Breiman's algorithm within the Random Forest technique to assess the importance of 16 input factors influencing the formation of landslide events in Son La province. The results obtained from this study serve as the foundation for selecting appropriate input factors to enhance the construction and accuracy of landslide susceptibility maps within the study area, especially in the context of climate change.