基于机器学习算法的县域台风灾害经济损失风险评估

RISK ASSESSMENT FOR TYPHOON ECONOMIC LOSSES IN COUNTY-BASED UNITS USING MACHINE LEARNING

关键词:  台风灾害  经济损失  风险评估  机器学习  随机森林

Key words:  typhoon disaster  economic losses  risk assessment  machine learning  random forest

摘要:

基于2008—2019年我国台风县(区)灾情的直接经济损失数据，根据经济损失率将台风灾害经济损失风险分为五类，考虑台风灾害的致灾因子和孕灾环境因子共选取10个解释变量，采用五种经典的机器学习算法，包括支持向量机(Support Vector Machine，SVM)、随机森林(Random Forest，RF)、AdaBoost、XGBoost(Extreme Gradient Boosting Machine)和LightGBM(Light Gradient Boosting Machine)，分别构建台风灾害经济损失风险评估模型，选出准确率最高的模型，应用于经典台风过程并进行检验评估。结果表明：基于RF算法的台风灾害经济损失风险模型的准确率最高；利用RF、XGBoost、LightGBM、AdaBoost和SVM算法构建模型的准确率依次为0.69、0.63、0.62、0.45和0.41。选择RF算法构建的台风灾害经济损失风险模型的解释变量表明，致灾因子是最主要的解释变量，其中，降雨导致损失的重要性超过风速。该模型在训练集和测试集上对风险分类的TS评分为0.55和0.51，但对每种风险类别的辨别能力存在差异，对于最低风险和最高风险的分类效果较好，对于较高风险和中等风险的分类能力不足。利用该模型对2017年第13号台风“天鸽”的经济损失进行检验评估，评估结果与实际台风灾害经济损失的风险等级较一致，各风险等级的准确率均达到0.7以上，TS评分在0.58以上，空报率和漏报率分别在0.31和0.25以下。

Abstract:

Based on the direct economic losses data of typhoon disasters in county-based units from 2008 to 2019, ten explanatory variables are selected by considering disaster-causing factors and environmental factors. Algorithms including Support Vector Machine (SVM), Random Forest (RF), AdaBoost, Extreme Gradient Boosting Machine (XGBoost) and Light Gradient Boosting Machine (LightGBM) are used to build typhoon economic losses risk models respectively. The performance is evaluated to identify the model with the best accuracy. The result show that the accuracy of typhoon economic losses risk model using RF algorithm is highest; the accuracy of models using RF, XGBoost, LightGBM, AdaBoost and SVM are 0.69, 0.63, 0.62, 0.45 and 0.41, respectively. According to the model based on RF, disaster-causing factors are the dominant explanatory variable, and the importance of rainfall-induced losses exceeded that of wind speed. The model is good at economic losses risk classification, with TS scores of 0.55 and 0.51 for the training and test sets. However, the model has different ability to distinguish between each category. The model has a good classification effect on the lowest risk and highest risk, but has a certain deficiency in higher risk and medium risk. This model is applied to Typhoon Hato, the 13th typhoon in 2017, and the results are consistent with the actual economic losses risk level, with the accuracy above 0.7, the TS score above 0.58, and the false alarm rate and false alarm rate below 0.31 and 0.25, respectively.