梁梓豪,苗鹏宇,Jianming Wang,王自法. 2024. 基于随机森林方法的地震损失预测. 地震学报,46(4):1−13. doi: 10.11939/jass.20220182
引用本文: 梁梓豪,苗鹏宇,Jianming Wang,王自法. 2024. 基于随机森林方法的地震损失预测. 地震学报,46(4):1−13. doi: 10.11939/jass.20220182
Liang Z H,Miao P Y,Jian M W,Wang Z F. 2024. Earthquake loss prediction based on random forest algorithm. Acta Seismologica Sinica46(4):1−13. doi: 10.11939/jass.20220182
Citation: Liang Z H,Miao P Y,Jian M W,Wang Z F. 2024. Earthquake loss prediction based on random forest algorithm. Acta Seismologica Sinica46(4):1−13. doi: 10.11939/jass.20220182

基于随机森林方法的地震损失预测

Earthquake loss prediction based on random forest algorithm

  • 摘要: 针对现有的基于实际震害评估大多研究仅限在某特定区域和某种结构类型,且所采用的数据样本数量也十分有限,本文基于随机森林模型,采用2011年3月11日东日本大地震的37万8 037条建筑物实际震害数据,利用美国应用技术协会(ATC-13)发布的地震震害等级划分预测了建筑物地震破坏所引起的损失,对建筑损失的影响因素进行了特征重要度分析。结果显示:通过SMOTE方法解决数据不均衡和贝叶斯优化超参数之后,得到了基于随机森林的预测模型测试集的准确率为68.8%,四种破坏等级的召回率分别为65.0%,53.6%,74.8%,81.8%;考虑生命安全性能将模型转换为二分类之后,模型准确率进一步提高至87.5%,极大地改善现有研究应用于建筑损失预测中数据样本数量受限、数据不均衡等导致的最严重破坏等级精度低等问题。对随机森林特征重要度的研究表明:震中距、PGA和vs30是模型输出影响最大的特征。通过本研究建立的地震损失评估模型能够实现快速且较为准确地预测地震引起的建筑破坏,可为震前规划和震后及时救援等提供参考。

     

    Abstract: Accurate assessment of earthquake damage is of crucial importance for pre-earthquake disaster prevention and mitigation, post-earthquake disaster relief and rapid reconstruction. Most of the existing studies based on actual earthquake damage assessment are limited to a specific region and a certain structure type, and the number of data samples used is also limited. Based on the random forest model, this paper uses 378,037 actual building damage data from the March 11, 2011 Japan Earthquake, and uses the earthquake damage classification issued by the American Applied Technical Council (ATC-13) to predict damage caused by earthquake damage to buildings and to analyze the feature importance of factors affecting building damage. The results show that after using SMOTE method to solve data imbalance and Bayesian approach to optimize of hyperparameters, the accuracy on the test set of the random forest-based prediction model is 68.8%, and the recall rates of the four damage classes are 65.0%, 53.6%, 74.8%, and 81.8%, respectively; the accuracy of the model is further increased to 87.5% by considering the life safety performance to convert the model to dichotomous classification, which greatly improves the existing research applied to construction loss prediction with limited number of data samples and low accuracy of the most severe damage level due to data imbalance. The study of the importance of random forest features showed that the epicenter distance, PGA and vs30 were the features with the greatest influence on the model output. The earthquake damage assessment model established by this study can achieve rapid and relatively accurate prediction of building damage caused by earthquakes, which is beneficial for the pre-earthquake planning and timely rescue after the earthquake.

     

/

返回文章
返回