Application of Transformer Models for Volcano Seismic Signals Classification

The detection and classification of volcanic seismic events are crucial for monitoring volcanic activity and minimizing natural disasters. Early identification of precursor seismic signals enables timely warnings, potentially saving lives in areas threatened by active volcanoes. We proposed an automatic seismic event classification method based on transformer models to maximize volcanic raw signal classification. The proposed method was trained and validated on the publicly available MicSigv1 dataset from the Cotopaxi volcano. The best model using a ten-fold cross-validation strategy achieved a mean of F1-score, accuracy, precision, recall, and area under the receiver operating characteristic curve scores of 0.959, 0.959, 0.960, 0.959, and 0.970 in the training phase and 0.965, 0.964, 0.966, 0.964, and 0.979 in the test phase, respectively. The successful classification performance in multiple metrics highlighted the effectiveness of the proposed method in capturing temporal dependencies from raw signals. This strength reduces the need for extensive feature extraction and complex preprocessing steps, improving the adaptability and scalability of the model for real-time monitoring systems.