Scholarlink

<p>The occurrence of landslides leads to deadly outcomes and annihilates infrastructure while disrupting affected communities. The implementation of effective prediction models serves to lower possible damage and enhance emergency readiness programs. To this end, in the present study, different machine learning algorithms, such as Random Forest (RF), Extreme Gradient Boosting (XGB), Gradient Boosting (GB), and Adaptive Boosting (Adaboost), are applied for the prediction of coverage areas in a landslide. Based on broad dataset availability in this regard, the performance of each algorithm is then evaluated using such important metrics as VAF, R<sup>2</sup>, MAE, MAPE, Maximum Error, and RMSE. The performance results indicate that among other models, Adaboost had the best performance for both training and testing with VAF of 99.99305% for training and 99.99795% for testing, MAE of 147.9712 for training and 256.0915 for testing, and MAPE of 3.14E-05 for training and 5.39E-05 for testing. Though the RF and GB models were found somewhat excellent in predictive performance, Adaboost was relatively finer in terms of its accuracy and reliability for landslide prediction as well as its risk management. This work emphasizes the further inclusion of machine learning models in arriving at more precise predictions of the landslides’ coverage area that can definitely contribute to better disaster preparedness and mitigation strategies.</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

Examining the role of machine learning methods for ANN prediction of landslide coverage area

  • 期刊论文

  • 发表日期：

    2025年6月2日

查看全文
  • Zhenhua Dai,
  • Weiguo Huang

摘要

The occurrence of landslides leads to deadly outcomes and annihilates infrastructure while disrupting affected communities. The implementation of effective prediction models serves to lower possible damage and enhance emergency readiness programs. To this end, in the present study, different machine learning algorithms, such as Random Forest (RF), Extreme Gradient Boosting (XGB), Gradient Boosting (GB), and Adaptive Boosting (Adaboost), are applied for the prediction of coverage areas in a landslide. Based on broad dataset availability in this regard, the performance of each algorithm is then evaluated using such important metrics as VAF, R2, MAE, MAPE, Maximum Error, and RMSE. The performance results indicate that among other models, Adaboost had the best performance for both training and testing with VAF of 99.99305% for training and 99.99795% for testing, MAE of 147.9712 for training and 256.0915 for testing, and MAPE of 3.14E-05 for training and 5.39E-05 for testing. Though the RF and GB models were found somewhat excellent in predictive performance, Adaboost was relatively finer in terms of its accuracy and reliability for landslide prediction as well as its risk management. This work emphasizes the further inclusion of machine learning models in arriving at more precise predictions of the landslides’ coverage area that can definitely contribute to better disaster preparedness and mitigation strategies.