杨雪雪, 刘强
中国海洋大学, 山东 青岛 266100

摘要:
作为破坏性最强的海洋灾害，风暴潮灾害每年都给我国沿海地区造成了巨大的经济损失，运用科学的方法模型合理预测风暴潮灾害经济损失对指导沿海地区的防灾减灾工作意义深远。本文基于风暴潮灾害的成灾特点建立了风暴潮灾害直接经济损失预评估指标体系，由于评估指标数据高度非线性，采用核主成分分析（KPCA）对高维非线性数据进行降维优化，并利用径向基函数（RBF）神经网络对降维后的数据进行训练，从而实现对风暴潮灾害直接经济损失的预测。选取广东省1996—2018年的32个风暴潮灾害损失样本对模型进行仿真测试，结果表明，KPCA-RBF预测模型集成了核主成分分析和径向基函数神经网络的优势，预测结果精度高，学习收敛速度快，对风暴潮灾害数据序列有较好的非线性拟合能力。
关键词:风暴潮经济损失预测核主成分分析(KPCA)RBF神经网络
DOI：10.11759/hykx20200322001
分类号:X43
基金项目:国家自然科学基金（41371496）；国家科技支撑计划项目（2013BAK05B04）

Economic loss assessment of storm-surge disasters based on the KPCA-RBF model
YANG Xue-xue, LIU Qiang
Engineering College, Ocean University of China, Qingdao 266100, China
Abstract:
As the most harmful marine disaster, storm-surge disasters cause serious economic losses to the coastal areas of China every year. It is crucial to reasonably assess the economic losses because of storm-surge disasters using scientific methods to guide the disaster prevention and mitigation work in coastal areas. Herein, based on the characteristics of storm-surge disasters, the assessment index system of the direct economic loss from a storm surge disaster is established. Owing to the storm=surge disaster, data loss is highly nonlinear; this study uses the Kernel Principal Component Analysis (KPCA) for nonlinear data dimension reduction optimization and the Radial Basis Function (RBF) neural network to train the dimension-reduced data to realize the assessment of direct economic loss owing to storm-surge disasters. This study collected data of 32 storm-surge disasters from 1996 to 2018 in the Guangdong Province to test the model. Results showed that the KPCA-RBF prediction model integrates the advantages of KPCA and the RBF neural network, which has high prediction accuracy, fast learning convergence speed, and good nonlinear fitting ability for storm-surge disaster data series.
Key words:storm surge disastereconomic loss assessmentkernel principal component analysisRBF neural network





PDF全文下载地址:

点我下载PDF