Economic loss assessment of typhoon disaster by integrated learning based on sparse self-coding
In order to eliminate the multi-collinearity between typhoon disaster data indicators and the over-fit-ting in the training process,this paper used sparse self-coding network to extract nonlinear features of typhoon disaster data in Guangdong province as the input of BP neural network.The output results of neural network were integrated by Bagging method to predict direct economic losses.In order to verify the advantage and effec-tiveness of sparse self-coding algorithm in feature extraction,the principal component model was introduced for comparative analysis.The experimental results showed that the sparse self-coding network could extract non-linear features and predict them,and its prediction accuracy is better than that of the linear dimension reduction method represented by principal component analysis.It also proved that the prediction effect of the integrated neural network proposed in this paper was better than that of the multi-layer BP neural network.It was proved that this method was an effective method to solve the problem of economic loss assessment of typhoon disas-ters,and had a certain application reference value.
NETL
NSTL
万方数据
