Imbalanced Seismic Event Discrimination Using Supervised Machine Learning

Abstract

The discrimination between earthquakes and artificial explosions is a significant issue in seismic analysis to efficiently prevent and respond to seismic events. However, the discrimination of seismic events is challenging due to the low incidence rate. Moreover, the similarity between earthquakes and artificial explosions with a local magnitude derives a nonlinear data distribution. To improve the discrimination accuracy, this paper proposes machine-learning-based seismic discrimination methods—support vector machine, naive Bayes, and logistic regression. Furthermore, to overcome the nonlinear separation problem, the kernel functions and regularized logistic regression are applied to design seismic classifiers. To efficiently design the classifier, P- and S-wave amplitude ratios on the time domain and spectral ratios on the frequency domain, which is converted by fast Fourier transform and short-time Fourier transform are selected as feature vectors. Furthermore, an adaptive synthetic sampling algorithm is adopted to enhance the classifier performance against the seismic data imbalance issue caused by the non-equivalent number of occurrences. The comparisons among classifiers are evaluated by the binary classification performance analysis methods.

Seismic Discrimination Between Nuclear Explosions and Natural Earthquakes using Multi-Machine Learning Techniques

“…On the other hand, the model obtained accuracy of 98% and 97.5% based on distinctive features, as reported by (Abdalzaher et al, 2021, andL. Dong et al, 2014) accuracy of 73% and 86% as reported by Ahn, H. et al, 2022 andAbdalzaher et al, 2021 in that order.…”

“…In the time domain waveform, the SVM model produced results with 66% accuracy, which translates to 99.4% accuracy based on feature extraction. According to study findings by Ahn, H. et al (2022), Kim et al (2020), Abdalzaher et al (2021), and L. Dong et al (2014), this model's accuracy was 93.7%, 95.5%, 86% and 96.3% respectively, based on distinguishing traits.…”

“…This is accomplished by adjusting each parameter for each model and then obtaining the best accuracy. These settings were chosen based on the best findings (Abdalzaher et Al., 2021;Ahn, h et al, 2022). The best parameters of the optimized best model for RF model has been achieved by 50 estimators, the criterion was 'entropy', and the,max_features is 'sqrt'.…”

“…To do this, fine-tune each model's parameter values until the optimal accuracy is achieved. The optimal results guided the selection of these settings (Abdalzaher et al, 2021;Ahn, H et al, 2022). We utilized ROC (receiver operating characteristic) and AUC (area under the curve) as evaluation metrics for classifier performance.…”

“…Several SVM algorithm types that have been proposed make use of feature vectors that are produced by estimating the frequency content in the frequency domain and calculating the amplitude from the time domain (Ahn, H. et al, 2022). Therefore, we used these machine learning techniques along with three other non-linear models (decision tree, K-nearest neighbors (KNN), and extreme gradient boosting) and one more linear model (logistic regression).…”