Landslide Susceptibility Prediction Based on Frequency Ratio Method and C5.0 Decision Tree Model

Abstract

This paper aims to propose an efficient landslide susceptibility prediction (LSP) model based on the frequency ratio method and C5.0 Decision Tree (C5.0 DT) model. Taking Ruijin City as the study area, local landslide inventory and 12 environmental factors are collected. Then the nonlinear correlations between landslide inventory and environmental factors are established by frequency ratio (FR) method. Thirdly, the FR values of these environmental factors are taken as the input variables of the C5.0 DT/SVM models; landslide samples and non-landslide samples are set as the output variables with values of 1 and 0, respectively. The mathematical relationship between input variables and output variables is established by C5.0 DT/SVM models. Finally, the performance of LSP of both models is evaluated by the Area Under Receiver Operation Characteristic Curve (AUC). Results show that: 1) The landslide susceptibility mapping (LSM) of the C5.0 DT and the SVM models are similar on the whole, high and very high susceptibility levels are mainly distributed in the north and in the edge of the study area. 2) The AUC values of C5.0 DT and SVM are 0.886 and 0.819, respectively. Both models have good LSP accuracy, however, the overall LSP accuracy of the C5.0 DT model is better than that of SVM. 3) It is significant and reliable to carry out LSP based on frequency ratio method and C5.0 DT model.

Modeling rules of regional flash flood susceptibility prediction using different machine learning models

“…Frontiers in Earth Science frontiersin.org (Sheng et al, 2022). Using the hydrology function of ArcGIS 10.2 Spatial Analyst Toolbox, streams in the study area are extracted from the DEM.…”

“…Flow analysis, depression determination, depression filling, confluence analysis, and river network analysis of DEM data by ArcGIS 10.2 software characterize hydrological environmental factors (Chang et al, 2022). The MNDWI is based on the normalized difference water index with a modification of the wavelength combinations that make up the index and is commonly used to represent surface hydrological information (Sheng et al, 2022). Using MNDWI, you can better reveal the microscopic characteristics of water bodies, such as the distribution of suspended sediments and changes in water quality, and identify water bodies in urban areas with high accuracy (Shu et al, 2022).…”

“…The study of a standardized process for constructing flash flood susceptibility prediction models using machine learning models, comparing and analyzing the prediction performance of different machine learning models, and carrying out practical engineering applications in Longnan County is of great practical importance for flash flood and geological hazard prevention and control, road planning, and development of appropriate risk mitigation strategies in the southwest and southern China (Sheng et al, 2022). Longnan County, in the southernmost part of Jiangxi Province, is used as an example in this study.…”

A Landslide Susceptibility Evaluation of Highway Disasters Based on the Frequency Ratio Coupling Model

“…In other work, Senem Tekin evaluated previous landslide characteristics and carried out landslide sensitivity modeling in the Ceyhan Watershed, using FR and LR, and the results provide a reference for future watershed management in the area [9]. Sheng Mingqiang et al applied the FR-SVM model for landslide susceptibility prediction, coupling the FR linkage method and SVM model, and compared the results with the single SVM model, showing that the FR-SVM model had a higher prediction accuracy [10].…”

Landslide Susceptibility Evaluation Based on Potential Disaster Identification and Ensemble Learning

“…The knowledge-driven LSE methods mainly include analytic hierarchy process (AHP) [ 5 , 25 , 26 ], fuzzy-AHP [ 27 ], fuzzy-relation AHP [ 28 ], fuzzy logic [ 29 , 30 ], fuzzy comprehensive evaluation [ 31 ], and fuzzy unordered rule induction [ 32 ] methods. The data-driven methods primarily consist of logistic regression [ 33 ], frequency ratio [ 34 ], weights of evidence [ 35 ], Information value [ 36 ], shallow machine learning (e.g., support vector machine [ 37 ], artificial neural network [ 38 ], random forest [ 39 ], and decision tree [ 40 ]), and deep learning (e.g., convolutional neural network [ 41 ], deep neural network [ 42 ], recurrent neural network [ 43 ], and deep belief network [ 44 ]) methods. Moreover, some ensemble methods were employed in LSE, including the combination of ant colony optimization and deep belief network [ 45 ], the ensemble of a radial basis function neural network, random subspace, attribute selected classifier, cascade generalization, and dagging [ 46 ], bagging based reduced error pruning trees [ 47 ] and so on.…”