Landslide hazard assessment based on Bayesian optimization–support vector machine in Nanping City, China

Abstract

Landslide hazard assessment is critical for preventing and mitigating landslide disasters. The tuning of model hyperparameters is of great importance to the accuracy and precision of one landslide hazard assessment model. In this study, Bayesian Optimization (BO) method was used to tune the hyperparameters of Support Vector Machine (SVM) model to obtain a high accuracy landslide hazard zoning map. 1711 historical landslide disaster points were obtained as landslide inventory in a case of Nanping City landslide hazard assessment. A total of 12 factors including elevation, slope, aspect, curvature, lithology, soil type, soil erosion, rainfall, river, land use, highway, and railway were selected as landslide conditional factors. The multicollinearity diagnosis was performed on factors using the Spearman correlation coe cient. 1711 landslides and 1711 non-landslides were collected as the dataset and divided into the same number of training dataset and testing dataset. The confusion matrix and receiver operating characteristic (ROC) curve were used to verify the models. The results of confusion matrix accuracy and the area under ROC curve (AUC) showed that BO-SVM (89.53%, 97%) performed better than only SVM (84.91%, 0.93), which indicated the superiority of the proposed method during this study.

Application of a Novel Hybrid Machine Learning Algorithm in Shallow Landslide Susceptibility Mapping in a Mountainous Area

“…We classified our landslides into two groups; training landslides (80%) and validation landslides (20%) (Xie et al, 2021a;Xie et al, 2021b). We then developed the hybrid RoFRF model and compared its performance to the four benchmark models including RF, ANN, BFT, and logistic model tree (LMT) using area under ROC and other statistical measures.…”

Automatic batch recognition of rock deformation areas based on image segmentation methods

“…proposed a CNN-to-FCN method to perform semantic segmentation on crack pixels in high-resolution images. Tavakkoli Piralilou et al (2019), Xie et al (2021) combined object-based images with multiple machine learning methods to conduct research.…”

Landslide susceptibility mapping (LSM) of Swat District, Hindu Kush Himalayan region of Pakistan, using GIS-based bivariate modeling

“…Landslide hazard is expected to grow due to anthropogenic interventions such as deforestation, population growth, urbanization, etc. and natural phenomena due to topographic, geologic, and climatic factors (Rahman et al, 2020;Xie et al, 2021aXie et al, , 2021bGuo et al, 2022;Ren et al, 2022;Wahla et al, 2022). Since the 20th century, mortality caused landslides has reached 6.2 million and has caused up to 10 billion US dollars of damage.…”