我国受地质灾害危害严重,采取区域预警是降低灾害带来损失的重要手段之一,其中预警模型是预警工作的关键。本文采用人工智能方法,基于历史灾害数据、地质环境数据及诱发因素数据,通过训练样本集构建、模型训练与优化评估等环节,构建基...我国受地质灾害危害严重,采取区域预警是降低灾害带来损失的重要手段之一,其中预警模型是预警工作的关键。本文采用人工智能方法,基于历史灾害数据、地质环境数据及诱发因素数据,通过训练样本集构建、模型训练与优化评估等环节,构建基于机器学习的滑坡灾害区域预警模型,模型准确率为0.934,AUC值为0.95,模型准确率及泛化能力均较好。通过实例校验,人工智能预警模型相较统计预警模型预警结果准确率有明显提高,预警结果的刻画也要更加精细,对滑坡灾害区域风险精细化防控有较好的指导作用。China is seriously affected by geological disasters. Taking regional early warning is one of the important means to reduce the losses caused by disasters, and the early warning model is the key to early warning work. This paper adopts artificial intelligence methods, based on historical disaster data, geological environment data and inducing factor data, through training sample set construction, model training and optimization evaluation, to construct a regional early warning model for landslide disasters based on machine learning. The model accuracy is 0.934, the AUC value is 0.95, and the model accuracy and generalization ability are good. Through example verification, the accuracy of the artificial intelligence early-warning model is obviously improved compared with the statistical early-warning model, and the description of the early-warning results should be more precise, it has a good guiding role in the refined prevention and control of regional risks of landslide disasters.展开更多
文摘我国受地质灾害危害严重,采取区域预警是降低灾害带来损失的重要手段之一,其中预警模型是预警工作的关键。本文采用人工智能方法,基于历史灾害数据、地质环境数据及诱发因素数据,通过训练样本集构建、模型训练与优化评估等环节,构建基于机器学习的滑坡灾害区域预警模型,模型准确率为0.934,AUC值为0.95,模型准确率及泛化能力均较好。通过实例校验,人工智能预警模型相较统计预警模型预警结果准确率有明显提高,预警结果的刻画也要更加精细,对滑坡灾害区域风险精细化防控有较好的指导作用。China is seriously affected by geological disasters. Taking regional early warning is one of the important means to reduce the losses caused by disasters, and the early warning model is the key to early warning work. This paper adopts artificial intelligence methods, based on historical disaster data, geological environment data and inducing factor data, through training sample set construction, model training and optimization evaluation, to construct a regional early warning model for landslide disasters based on machine learning. The model accuracy is 0.934, the AUC value is 0.95, and the model accuracy and generalization ability are good. Through example verification, the accuracy of the artificial intelligence early-warning model is obviously improved compared with the statistical early-warning model, and the description of the early-warning results should be more precise, it has a good guiding role in the refined prevention and control of regional risks of landslide disasters.
