摘要
随着国家整体防灾减灾能力的提升,洪涝灾害损失呈逐年波动下降趋势,但单次灾害损失波动幅度巨大,在遭遇农业大灾风险时,保险业承担能力有限。财政补贴与保险费率厘定需基于预期损失的精准评估,而基于传统灾害损失的评估精度较低,运用风险序贯性链式结构理论,结合机器学习提高损失评估精度。以河北省玉米洪涝损失评估为例,检验发现,以决策树、BP神经网络、XGBoost回归模型预测区域降水量表征的风险分级函数对SVM模型加以改进,显著提高了灾害损失评估模型精度,而随着国家防灾减灾整体能力的提升,灾害损失呈现波动下降趋势。
With the improvement of China’s overall disaster prevention and mitigation capacity,the loss of flood disaster fluctuates year by year,but the fluctuation range of single disaster loss is huge,and the insurance industry has limited ability to bear the risk of agricultural catastrophe.The determination of financial subsidies and insurance rates requires accurate assessment based on expected losses,which is usually based on traditional disaster losses.The theory of risk sequential chain structure and machine learning are used to improve the accuracy of loss assessment.Through taking the evaluation of corn flood loss in Hebei Province as an example,it is found that the decision tree,BP neural network and XGBoost regression model are used to predict the risk classification function of regional precipitation,and the SVM model is improved,which significantly improves the accuracy of disaster loss assessment model.Moreover,with the improvement of the overall ability of national disaster prevention and mitigation,disaster losses show a downward trend.
作者
李炳萱
蒲成毅
Li Bingxuan;Pu Chengyi(School of Insurance,Central University of Finance and Economics,Beijing 102206,China)
出处
《黑龙江科学》
2024年第16期63-67,共5页
Heilongjiang Science
基金
科技创新2030“新一代人工智能”重大课题“不确定环境下农业大灾风险转移”(2022ZD0119504)。
作者简介
李炳萱(2002-),男,本科生。研究方向:保险与风险管理。
