基于改进神经网络及地理信息系统空间分析的风暴潮经济损失评估

台风风暴潮是影响我国最严重的海洋灾害之一,台风风暴潮经济损失的预评估对防灾减灾有重要作用。针对目前大量单机器学习模型评估效果不佳的问题,提出基于天牛须搜索算法优化的BP神经网络及GIS空间分析的台风风暴潮经济损失评估模型。该模型在神经网络训练前,先利用天牛须搜索算法对神经网络的初始权值阈值进行优化,以提高网络的寻优精度；在网络输出计算结果后,借助GIS空间分析功能找出待评估台风风暴潮的相似样本,并用相似样本对待评估台风风暴潮进行结果调整；最后基于悲观、乐观估计思想,给出台风风暴潮经济损失预评估区间。计算结果表明,结合GIS空间分析调整能给出更有效的模型评估结果。

The assessment of storm surge disaster loss based on improved neural network and GIS spatial analysis

Typhoon storm surge is one of the most serious marine disasters affecting China. The pre-assessment of the economic loss caused by typhoon storm surge plays an important role in disaster prevention and reduction. In view of the poor evaluation effect of a large number of single machine learning models, a typhoon storm surge economic loss evaluation model based on BAS(Beetle Antennae Search algorithm)-BP neural network and GIS spatial analysis is proposed in this paper. This model optimizes the initial weight of neural network using the BAS algorithm to improve the optimization accuracy before training neural network. After the model output results, GIS spatial analysis function is used to find out similar samples of typhoon storm surge to be evaluated, and similar samples are used to adjust the results of typhoon storm surge evaluation. Finally, based on the pessimistic and optimistic ideas, the forecast range of the economic loss of typhoon storm surge is given. The results show that the model evaluation results can be more effectively obtained by combining GIS spatial analysis and adjustment.