Speaker
Description
Discrimination between explosions and earthquakes is a major challenge in the field of seismology. This task is important not only to meet the expectations of the Comprehensive Nuclear Test Ban Treaty (CTBT) but also to refine seismic bulletins used in regional seismicity research, seismotectonic analysis, and seismic hazard assessment.
Seven models using different machine learning algorithms (logistic regression, support vector machine - SVM, K-nearest neighbours, decision tree, random forest, and Naive Bayes) were trained by selecting all specific features in REB bulletins and acceptable performance metrics were obtained in the test phase. However, after performing a feature selection analysis, removing irrelevant and redundant features, and adding features from the waveform without parametric changes in the models, the performance metric increased significantly.
Meaningful evaluation metrics were used to assess the machine learning models, including accuracy, precision, recall (sensitivity), F1 score and ROC curve.
| [email protected] |
