Multi-class earthquake magnitude classification in Kahramanmaraş region using feature engineering: artificial intelligence approaches

This study presents a multi-class earthquake magnitude classification framework based on 39,849 earthquake records from the Kahramanmara & scedil; region covering 1994-2025. Earthquake magnitudes are categorized into six classes based on natural statistical distributions in the dataset. A total of 32 features capturing seismic activity, time series dynamics, geophysical properties, and statistical metrics are extracted and further enriched through physical and temporal analyses. Multiple machine learning and deep learning models, including Random Forest, Gradient Boosting, Logistic Regression, Support Vector Machine (SVM), Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), and one-dimensional Convolutional Neural Network (1D-CNN), are trained and evaluated using accuracy, precision, sensitivity, specificity, F1-score, and cross-validation metrics. Gradient Boosting achieves the highest performance with 98.70% accuracy and a 98.71% F1-score. Logistic Regression emerges as a strong alternative with 96.43% accuracy, while ANN attains the highest performance among deep learning models with 83.88% accuracy. Feature engineering, particularly the integration of seismic energy indicators, temporal dependency measures, and geophysical parameters, was found to substantially enhance classification accuracy. The proposed approach combines multi-class classification with systematic feature engineering, providing a robust methodological basis for earthquake magnitude prediction and offering potential applications in disaster management, early warning systems, and seismic risk assessment.

Dergi

NATURAL HAZARDS

Yıl

2026

Cilt

122

Sayı

6

ISSN

0921-030X

eISSN

1573-0840

DOI

10.1007/s11069-026-08011-1

WoS UID

WOS:001703085100003