Scholarlink

<p>Improving the spatial representativeness of non-landslide samples remains a major challenge for enhancing the stability and predictive accuracy of landslide susceptibility assessment (LSA).To address this issue, this study investigated the Longyang Gorge–Liujiaxia section in the upper Yellow River Basin, where eleven conditioning factors were used to establish the susceptibility evaluation system. To enhance the spatial representativeness of non-landslide samples, a density-based spatial clustering of applications with noise (DBSCAN) algorithm was utilized to eliminate spatially anomalous samples, in combination with the synthetic minority oversampling technique (SMOTE) to balance the dataset. Model predictive performance and uncertainty were assessed using the area under the curve (AUC) of receiver operating characteristic (ROC) and the variation characteristics of the susceptibility index distribution. The results showed that the DBSCAN-SMOTE sampling strategy has the highest evaluation accuracy. Compared with the random generation method, the AUC value of the support vector machine model (SVM) and logistic regression model (LR) based on this sampling strategy were improved by 46.98% and 35.71% respectively. Overall, the proposed non-landslide sampling strategy significantly improves sample quality and model robustness, providing a practical and generalizable solution for an accurate and reliable LSA.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Landslide susceptibility assessment under non-landslide sampling strategy based on a clustering algorithm

  • 期刊论文

  • 发表日期：

    2026年2月10日

查看全文
  • Jinhao Meng,
  • Juan Wang,
  • Benbo Sun,
  • Juhong Han,
  • Chengfang Huang

摘要

Improving the spatial representativeness of non-landslide samples remains a major challenge for enhancing the stability and predictive accuracy of landslide susceptibility assessment (LSA).To address this issue, this study investigated the Longyang Gorge–Liujiaxia section in the upper Yellow River Basin, where eleven conditioning factors were used to establish the susceptibility evaluation system. To enhance the spatial representativeness of non-landslide samples, a density-based spatial clustering of applications with noise (DBSCAN) algorithm was utilized to eliminate spatially anomalous samples, in combination with the synthetic minority oversampling technique (SMOTE) to balance the dataset. Model predictive performance and uncertainty were assessed using the area under the curve (AUC) of receiver operating characteristic (ROC) and the variation characteristics of the susceptibility index distribution. The results showed that the DBSCAN-SMOTE sampling strategy has the highest evaluation accuracy. Compared with the random generation method, the AUC value of the support vector machine model (SVM) and logistic regression model (LR) based on this sampling strategy were improved by 46.98% and 35.71% respectively. Overall, the proposed non-landslide sampling strategy significantly improves sample quality and model robustness, providing a practical and generalizable solution for an accurate and reliable LSA.