ML backazimuth estimation for earthquakes in the Irpinia region

This dataset summarizes the experimental results obtained during the Transnational Access C1_TA3-84-1_2, conducted within the framework of the Geo-INQUIRE project. We used machine learning models trained on a large-scale global dataset to estimate back-azimuth at individual three-component seismic waveforms at the Irpinia Near-Fault Observatory (Southern Italy). We consider two representative architectures: eXtreme Gradient Boosting (XGB) and Bayesian Neural Networks (BNN). Model performance is benchmarked against the classical Principal Component Analysis (PCA) and Adjustable Time Window–PCA (ATW-PCA) model. Seismic waveforms can be accessed at http://isnet-bulletin.fisica.unina.it/cgi-bin/isnet-events/isnet.cgi?