“…With the development of computing power, geospatial data, and technologies, a growing number of studies have explored the potential of using machine learning (ML) based algorithms with geographic information system (GIS) datasets and remote sensing images for constructing landslide susceptibility models on the regional scale, such as decision trees (DT) [8,9,10], entropy- and evolution-based algorithms [11,12,13,14], fuzzy theory [15,16,17], neural network [12,18,19,20,21,22], neural-fuzzy systems [23,24], and support vector machines [20,21]. Random forests (RF) has received increased attention in the ML domain in recent years because of the following advantages: (1) Excellent accuracy [25], (2) processing speed [26,27], (3) few parameter settings [28,29], (4) availability of high-dimensional data analysis (e.g., hyperspectral image cubes) [30,31], and (5) insensitivity to imbalanced training data [32].…”