Regression and Neural Network Modeling of Earthquake Hypocenter Depth in the Broader Croatia–Adriatic–Dinarides Region
Proceedings of reviewed papers: 6th Croatian geomathematical congress · 03.08.2026 · 6 · 127-139
DOI: 10.17794/cgc.2026.9 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of reviewed papers: 6th Croatian geomathematical congress, online erschienen am 03.08.2026. Verfasst von Toni Čvrljak.
Abstract (Originalsprache)
This study investigates the application of regression methods and neural networks to model earthquake depth across the broader Croatia–Adriatic–Dinarides region. Publicly available data from the United States Geological Survey (USGS) earthquake catalog for the period 2000–2025 were analyzed, comprising 6,197 seismic events of magnitude 2.0 or greater. Epicentral latitude and longitude, magnitude, year, and calendar month of occurrence were used as input features, while hypocenter depth was defined as the target variable. A Dummy Regressor, linear regression, a multilayer perceptron, and a radial basis function neural network were compared. The models were evaluated using five-fold cross-validation based on mean absolute error, mean squared error, and root mean squared error. The multilayer perceptron achieved the best predictive performance, with an average MAE of 5.58 ± 0.88 km and an RMSE of 18.10 ± 4.03 km. Compared with the reference Dummy Regressor, the multilayer perceptron reduced the average RMSE by approximately 25.7%, indicating the advantage of a nonlinear model in describing the relationships between the spatial, temporal, and magnitude-related characteristics of earthquakes and hypocenter depth.
Zitieren
APA
Čvrljak, T. (2026). Regression and Neural Network Modeling of Earthquake Hypocenter Depth in the Broader Croatia–Adriatic–Dinarides Region. Proceedings of reviewed papers: 6th Croatian geomathematical congress, 6, 127-139. https://doi.org/10.17794/cgc.2026.9
Vancouver
Čvrljak T. Regression and Neural Network Modeling of Earthquake Hypocenter Depth in the Broader Croatia–Adriatic–Dinarides Region. Proceedings of reviewed papers: 6th Croatian geomathematical congress. 2026;6:127-139. doi:10.17794/cgc.2026.9
BibTeX
@article{vrljak2026,
author = {Čvrljak, Toni},
title = {Regression and Neural Network Modeling of Earthquake Hypocenter Depth in the Broader Croatia–Adriatic–Dinarides Region},
journal = {Proceedings of reviewed papers: 6th Croatian geomathematical congress},
year = {2026},
volume = {6},
pages = {127--139},
doi = {10.17794/cgc.2026.9}
}
EUVIMED