Scholarlink

<p>Taba Penanjung area is one of the areas prone to landslides in Bengkulu Province, Indonesia. In this research, we aim to investigate Landslide Susceptibility Map (LSM) at region using Machine Learning (ML) techniques such as Frequency Ratio (FR), Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGB). The fourteen parameters, likely referring to factors influencing landslide susceptibility, have been utilized. We found that the LSM can be divided into three landslide-prone zones, namely low, moderate, and high, with areas of 21.50%, 37.54% and 40.97% respectively. Our results showed that the LR, RF, SVM, and XGB models can generate models with accuracy of 88.46%, 85.90%, 95.51%, and 82.05%, respectively. Many ML models have achieved an Area Under Curve (AUC) value &gt; 0.82, indicating high reliability of the susceptibility map generated even; the SVM model provides the AUC of &gt; 0.95 with the best performance of the LSM prediction. Our study demonstrates the effectiveness of ML models in analyzing the LSM, but in the future, effective landslide risk mitigation in this region must incorporate detailed real-time landslide monitoring incorporation with community-based awareness programs, alongside the enforcement of land-use regulations and sustainable engineering practices.</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 mapping along Taba Penanjung area, Bengkulu province, Indonesia using machine learning (ML) techniques

  • 期刊论文

  • 发表日期：

    2025年8月22日

查看全文
  • Surfiarti Maharani,
  • Ashar Muda Lubis,
  • Elfi Yuliza,
  • Rida Samdara,
  • Bambang Setiadi

摘要

Taba Penanjung area is one of the areas prone to landslides in Bengkulu Province, Indonesia. In this research, we aim to investigate Landslide Susceptibility Map (LSM) at region using Machine Learning (ML) techniques such as Frequency Ratio (FR), Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGB). The fourteen parameters, likely referring to factors influencing landslide susceptibility, have been utilized. We found that the LSM can be divided into three landslide-prone zones, namely low, moderate, and high, with areas of 21.50%, 37.54% and 40.97% respectively. Our results showed that the LR, RF, SVM, and XGB models can generate models with accuracy of 88.46%, 85.90%, 95.51%, and 82.05%, respectively. Many ML models have achieved an Area Under Curve (AUC) value > 0.82, indicating high reliability of the susceptibility map generated even; the SVM model provides the AUC of > 0.95 with the best performance of the LSM prediction. Our study demonstrates the effectiveness of ML models in analyzing the LSM, but in the future, effective landslide risk mitigation in this region must incorporate detailed real-time landslide monitoring incorporation with community-based awareness programs, alongside the enforcement of land-use regulations and sustainable engineering practices.