论文部分内容阅读
风暴潮灾害是影响我国最严重的海洋灾害,风暴潮灾害损失的预评估对防灾减灾有重要作用。本文选用2002~2014年的40组风暴潮历史灾情资料进行试验,首先建立风暴潮灾害损失评估指标体系并用灰色关联分析法对指标进行筛选,然后采用最优权重组合将支持向量机和BP神经网络进行组合预测分别对风暴潮直接经济损失和受灾人口数进行预测,并与单一预测方法进行对比,发现组合预测方法可以降低误差,提高损失预测的准确性,建立风暴潮灾害损失预评估模型,为决策者进行预警信息的发布提供有效依据。
Storm surge disaster is the most serious marine disaster in our country. The pre-assessment of storm surge damage plays an important role in disaster prevention and mitigation. In this paper, 40 groups of storm surge history data from 2002 to 2014 are selected for testing. First, the assessment index system of storm surge damage is established, and the index is screened by gray relational analysis. Then the optimal weight combination is used to combine SVM and BP neural network Combining forecast with direct economic loss and population affected by storm surge, respectively, and compared with the single prediction method, we find that combined forecasting method can reduce the error and improve the accuracy of loss prediction, and establish the pre-storm damage assessment model as Decision makers to provide early warning information to provide an effective basis for the release.