A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
A frequency ratio–based sampling strategy for landslide susceptibility assessment
Abstract Owing to the difficulties in determining the boundaries of landslides from landslide inventory, landslide samples are often expressed by geometrical points, rather than the actual shape of landslides. However, when grid units are used for landslide susceptibility assessment (LSA), the number of landslide samples is quite limited by using geometric points to represent landslides. This situation further leads to a small number of non-landslide samples that are randomly generated based on the landslide samples. This paper proposes a frequency ratio (FR)-based sampling strategy to enhance the information of landslides to achieve an improved machine learning-based LSA. To realize this idea, the FR values of landslide conditioning factors are first calculated, and each grid unit has an attribute dataset composed of FR values. Then, the grid units with the minimum FR value no less than 1 are selected as supplementary positive samples, the grid units with the maximum FR value less than 1 are selected as supplementary negative samples. The enhanced dataset is finally composed of supplementary positive samples and historical landslide points. Thereafter, two typical ML models of the random forest (RF) and support vector machine (SVM) are employed to construct the improved LSA models based on the enhanced datasets. Finally, the landslide susceptibility of Anhua County, China, is assessed by the improved ML models, and the results are compared with the ML models trained by the traditional sampling method. The results indicate that compared with traditional RF and SVM models, the corresponding improved models show better performance in accuracy, precision, recall, and F1 value. In particular, the AUC values of the two models are increased by 9% and 8%, respectively. The proposed method provides a promising alternate for an accurate and reliable LSA.
A frequency ratio–based sampling strategy for landslide susceptibility assessment
Abstract Owing to the difficulties in determining the boundaries of landslides from landslide inventory, landslide samples are often expressed by geometrical points, rather than the actual shape of landslides. However, when grid units are used for landslide susceptibility assessment (LSA), the number of landslide samples is quite limited by using geometric points to represent landslides. This situation further leads to a small number of non-landslide samples that are randomly generated based on the landslide samples. This paper proposes a frequency ratio (FR)-based sampling strategy to enhance the information of landslides to achieve an improved machine learning-based LSA. To realize this idea, the FR values of landslide conditioning factors are first calculated, and each grid unit has an attribute dataset composed of FR values. Then, the grid units with the minimum FR value no less than 1 are selected as supplementary positive samples, the grid units with the maximum FR value less than 1 are selected as supplementary negative samples. The enhanced dataset is finally composed of supplementary positive samples and historical landslide points. Thereafter, two typical ML models of the random forest (RF) and support vector machine (SVM) are employed to construct the improved LSA models based on the enhanced datasets. Finally, the landslide susceptibility of Anhua County, China, is assessed by the improved ML models, and the results are compared with the ML models trained by the traditional sampling method. The results indicate that compared with traditional RF and SVM models, the corresponding improved models show better performance in accuracy, precision, recall, and F1 value. In particular, the AUC values of the two models are increased by 9% and 8%, respectively. The proposed method provides a promising alternate for an accurate and reliable LSA.
A frequency ratio–based sampling strategy for landslide susceptibility assessment
Liu, Lei-Lei (author) / Zhang, Yi-Li (author) / Xiao, Ting (author) / Yang, Can (author)
2022
Article (Journal)
Electronic Resource
English
BKL:
56.00$jBauwesen: Allgemeines
/
38.58
Geomechanik
/
38.58$jGeomechanik
/
56.20
Ingenieurgeologie, Bodenmechanik
/
56.00
Bauwesen: Allgemeines
/
56.20$jIngenieurgeologie$jBodenmechanik
RVK:
ELIB18
A modified frequency ratio method for landslide susceptibility assessment
| British Library Online Contents | 2017