Identification of earthquake-prone areas in the Tashkent region using machine learning algorithms

Jasur Mamarakhimov

E3S Web of Conferences, 2026

Full text Open access: Gold Metadata (CSL-JSON)
DOI 10.1051/e3sconf/202674501001
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

Identifying earthquake-prone areas is a key component of seismic hazard assessment in tectonically active regions. This study developed a comprehensive geospatial database integrating geological, morphostructural, tectonic, seismological, and geodynamic parameters for identifying earthquake-prone areas in the Tashkent region. Unlike conventional approaches, numerical stress-strain modeling results of the Earth's crust were incorporated into machine learning as physically meaningful input features. The study area was divided into spatial zones, each represented by 29 binary predictor features describing its geological, tectonic, seismological, and geodynamic characteristics, while earthquake occurrence within the target magnitude ranges was used as the classification target. The Tashkent region was excluded from model training and used as an independent spatial test area. Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, Support Vector Machine, K-Nearest Neighbors, XGBoost, Neural Network, and Kora-4 were evaluated using Accuracy, Precision, Recall, F1-score, and ROC-AUC. Under highly imbalanced class conditions, Kora-4 demonstrated the strongest overall performance, achieving Accuracy/F1-score values of 0.80/0.80 for 4≤M<5 and 0.82/0.85 for 5≤M<6, respectively. The results indicate that integrating geodynamic modeling with geological, tectonic, and seismological parameters provides physically meaningful predictors for reliable earthquake-prone area identification and regional seismic hazard assessment.