Automatic Identification of Seismic Signal Based on Entropy Feature and Genetic Algorithm Optimized BP Neural Network

In order to improve the accuracy of seismic event identification, a new method of applying distribution entropy, fuzzy entropy, and sample entropy to identify seismic waves is proposed, and the back propagation(BP) neural network optimized by genetic algorithm (GA) is used to test the identification effect. The seismic event identification effect is measured based on AUC, recall, accuracy, specificity and other indicators by applying some waveform records of the 2021 Qinghai Mado Ms7.4 earthquake and Yunnan Yangbi Ms6.4 and Ms5.6 earthquake. The results show that the combined criterion feature is better than a single criterion in the recognition effect. For a single criterion, the distinguishing ability of sample entropy is the best; the overall recognition effect of GA-BP is better than Boosting, decision tree, naive Bayes, et al. Finally, GA-BP model can be considered as an automatic identification method for identifying seismic waveform.