Landslide Susceptibility Modeling Using Remote Sensing Data and Random SubSpace-Based Functional Tree Classifier

In this study, a random subspace-based function tree (RSFT) was developed for landslide susceptibility modeling, and by comparing with bagging-based (BFT), classification regression (CART), Naïve-Bayes (NBTree) Classifier, to judge the performance difference between hybrid model single models. first step, according characteristics of geological environment previous literature, 12 conditioning factors were selected, including aspect, slope, profile curvature, plan elevation, topographic wetness index (TWI), lithology, normalized vegetation (NDVI), land use, soil, distance river road. Secondly, 328 historical landslides randomly divided into training group validation in ratio 70/30, important analysis points conditional carried out using functional (FT) model. third all data are loaded FT, RSFT, BFT, CART, NBTree models generation maps (LSM). Comparisons made area under receiver operating characteristic curve (AUC) determine efficiency effectiveness. According verification results, five selected time perform reasonably, but RSFT has highest prediction rate (AUC = 0.838), which is better than other three machine learning The results study also demonstrated that generally improves predictive power benchmark