Identification of Earthquake Source Attributes Based on DAPSO-BP Combined Model

Earthquake is a complex phenomenon of crustal movement and countless seismic disasters occur every year all over the world. Therefore, in order to reduce the impact of seismic disasters, the identification of natural earthquakes and artificial blasting needs to be studied. We propose a combined deterministic and adaptive dragonfly algorithm and particle swarm optimization and back propagation (DAPSO-BP) model to achieve the distinction between natural earthquakes and artificial blasting. The hyperparameters of back propagation (BP) neural network were optimized by DAPSO algorithm to improve the training effect of BP neural network model. Firstly, the raw data were cleaned to eliminate invalid values and outliers. To address the problem, the use of sample entropy was considered as the feature input of the model. The original signal was first decomposed into several intrinsic mode functions (IMF) by empirical mode decomposition (EMD) decomposition algorithm, and then the sample entropy was calculated for the IMF components, and then the sample entropy feature dataset was constructed. Then, it is inputted into the DAPSO-BP combined model to complete the identification of natural earthquakes and artificial blasting. Finally, the recognition results were compared with other models (BP neural network, support vector machine), and our proposed DAPSO-BP combination model had an $$R^{2}$$ of 99.4 $$\%$$ , the mean absolute error (MAE) of 0.014, and the root mean square error (RMSE) of 0.023, which is the best recognition effect among all models.