基于信息融合的自然灾害等级评估方法研究

为对自然灾害灾情等级进行准确评估,在BP神经网络模型的基础上,结合DS证据理论建立基于信息融合的自然灾害灾情等级评估模型。该模型通过对输入的灾害评估指标数据进行分类,建立网络组,对网络组的输出,建立对于各类信任度的基本概率分配函数,最后利用DS证据理论融合,从而实现灾害的最终等级评估。在MATLAB环境下,以我国45个自然灾害的灾情历史资料数据为训练样本进行模型训练,并对2009年自然灾害灾情进行评估测试。结果表明,该模型能改善单一BP神经网络不稳定、误差大的缺点,得到较优的结果。

Research on Assessment Method for Natural Disaster Loss Grade Based on Information Fusion

To more accurately assess the grade of natural disasters,an information fusion assessment model for natural disaster grade has been built based on BP neural network and DS evidential theory.In this model,the input index data of natural disasters are firstly classified and the BP neural networks groups are constructed.And then the basic probability assignment functions with different confidence level are obtained according to the network group's outputs.Finally,a fusion is made with DS evidential theory and the final grade of natural disasters is obtained.Under MATLAB environment,the model is trained with the data of 45 historical natural disaster cases in China as the training samples,and tests are conducted on the model with the data of natural disasters in 2009.The results show that this model can obtain better assessment result by solving the problems of unstable results and poor precision of single BP neural network.