基于GAN的地震人员死亡样本扩充方法研究OA北大核心CSTPCD

Earthquake is a natural disaster with great destructiveness,which seriously threatens the social and eco-nomic development.Accurate prediction of the number of deaths at the first time after an earthquake is of great significance for disaster relief and emergency response.However,due to the limited number of historical earthquake samples,it is diffi-cult to ensure the stability of small sample models.In this paper,GAN is used to expand the samples to obtain an augment-ed dataset that is highly similar to the original dataset.The empirical model is then fitted using the two datasets,and the training and prediction performance of the machine learning models(SVR,ELM,BP,DNN)are also compared.The re-sults show that the augmented dataset not only enhances the fitting effect of the empirical models,but also has a significant effect on the prediction performance of the machine learning models,and the root-mean-square error between the predicted and true values of each model is reduced by an average of 48.37%after the sample size expansion compared to the pre-expansion period.Therefore,the augmented dataset obtained by GAN expansion can be used as an effective supplement to the original samples,and is an effective way to improve the accuracy of the earthquake personnel death model.