基于随机权神经网络的地震灾害经济损失评估与预测  

Earthquake Disaster Economic Loss Estimation and Prediction Based on Neural Networks with Random Weights

作　　者：谢家智[1] 车四方 林涌[1] 

机构地区：[1]西南大学经济管理学院,重庆北碚400715

出　　处：《灾害学》2017年第1期1-4,10,共5页Journal of Catastrophology

基　　金：教育部规划基金项目(12YJA790149);国家社科基金重点项目(12AGL008)

摘　　要：地震灾害损失的评估和预测是地震风险管理的重要环节。利用随机权神经网络(NNRW)对我国2008-2014年地震灾害直接经济损失进行了评估和预测,并将其与传统的BP神经网络进行了比较。研究结果表明,在训练时间、训练精度和预测精度上随机权神经网络都优于BP神经网络,并且随机权神经网络不易发生过拟合现象。因此,随机权神经网络为地震灾害直接经济损失的快速而精准评估和预测提供了一种新方法。The estimation and prediction of earthquake disaster loss is an important part of seismic risk management. The earthquake disaster direct economic loss of 2008 -2014 in China was evaluated by using the NNRW, and compared with the traditional BP neural network. Experimental results show that the NNRW is better than the BP neural network in training time, training accuracy and prediction accuracy. Therefore, the NNRW provides a new method for the rapid and accurate assessment and prediction of direct economic loss of earthquake disaster.

关 键 词：随机权神经网络 地震灾害 直接经济损失 评估 预测 

分 类 号：X43[环境科学与工程—灾害防治]