Scholarlink

<p>Landslide activity classification is crucial for disaster risk management and mitigation. This study explores the effectiveness of ensemble learning algorithms in classifying landslide activities using Vegetation Anomalies Indicator (VAI) within Kundasang, Sabah, Malaysia. Seven groups of VAIs were selected based on their relevance to the geo-environmental conditions of the study area. The analysis incorporated a comprehensive dataset of landslide occurrences, with 70% of the sites designated for training and 30% for validation. Ensemble learning algorithms, specifically boosting and bagging were employed to classify landslide activity. The performance of these algorithms was evaluated against a landslide inventory map using several metrics: producer accuracy (PA), user accuracy (UA), overall accuracy (OA), and kappa coefficient (κ). Results demonstrate that Decision Tree (DT) and Stochastic Gradient Boosting (SGB) achieved the highest OA of 80.0% and 75.9% at a 1-m resolution, respectively. For bagging algorithms, Random Forest (RF) outperformed BAGGED CART with an OA of 81.9% at a 1-m resolution. Additionally, the κ indicated a substantial agreement for both ensemble methods, with values reaching up to 0.728 for RF. These findings underscore the potential of utilizing VAIs combined with ensemble learning for effective landslide activity classification, contributing valuable insights for improving landslide risk assessment and management 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

A comparative study of ensemble learning algorithms for the classification of landslide activity using vegetation anomalies indicator (VAI): a case study of Kundasang, Sabah

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月10日

查看全文
  • Mohd Radhie Mohd Salleh,
  • Muhammad Zulkarnain Abdul Rahman,
  • Zamri Ismail,
  • Mohd Faisal Abdul Khanan,
  • Radzuan Sa’ari,
  • Ahmad Razali Yusoff

摘要

Landslide activity classification is crucial for disaster risk management and mitigation. This study explores the effectiveness of ensemble learning algorithms in classifying landslide activities using Vegetation Anomalies Indicator (VAI) within Kundasang, Sabah, Malaysia. Seven groups of VAIs were selected based on their relevance to the geo-environmental conditions of the study area. The analysis incorporated a comprehensive dataset of landslide occurrences, with 70% of the sites designated for training and 30% for validation. Ensemble learning algorithms, specifically boosting and bagging were employed to classify landslide activity. The performance of these algorithms was evaluated against a landslide inventory map using several metrics: producer accuracy (PA), user accuracy (UA), overall accuracy (OA), and kappa coefficient (κ). Results demonstrate that Decision Tree (DT) and Stochastic Gradient Boosting (SGB) achieved the highest OA of 80.0% and 75.9% at a 1-m resolution, respectively. For bagging algorithms, Random Forest (RF) outperformed BAGGED CART with an OA of 81.9% at a 1-m resolution. Additionally, the κ indicated a substantial agreement for both ensemble methods, with values reaching up to 0.728 for RF. These findings underscore the potential of utilizing VAIs combined with ensemble learning for effective landslide activity classification, contributing valuable insights for improving landslide risk assessment and management practices.