Analysis of Optimal Buffer Distance for Linear Hazard Factors in Landslide Susceptibility Prediction

A linear hazard-causing factor is the environmental element of landslide susceptibility prediction, and setting buffer distance a has an important influence on accuracy prediction based machine learning algorithms. geographic information system (GIS) generally been accepted in correlation analysis between factors landslides; most common are statistical models zone superposition for causative distances counts. However, there problem process model building: that used to build its results can appropriately reflect disaster-causing landslides. To solve this problem, density under different single-loop was established Pearson’s method with 12 factors, such as elevation, topographic relief, from water road, Ruijin City, Jiangxi Province obtain relevant combinations; random forest (RF) were then predict susceptibility. Finally, Kappa coefficient distribution characteristics index investigate modeling laws. The indicate correlated combination reaches 96.65%, error rate non-landslide points 4.2%, 11.3%, which higher than same combination, confirming reasonableness using distance.