Accurate Prediction of Peak Ground Acceleration Using Random Forests: A Data-Driven Approach with PEER Updated NGA-WEST2 Ground Motion Records

Abstract

Accurate prediction of peak ground acceleration (PGA) is crucial for seismic hazard assessment and earthquake-resistant design.Traditional regression-based ground motion prediction equations often fall short in capturing the complex, nonlinear interactions among earthquake parameters. This study proposes a machine learning approach using a random forest (RF) model to predict PGA based on five key input variables: moment magnitude (Mw), closest distance to the rupture plane (ClstD), hypocentral depth, rake angle, and average shear wave velocity in the top 30 meters (Vs30). A comprehensive dataset of 16,211 ground motion records from the Pacific Earthquake Engineering Research Center (PEER), Updated NGA-WEST2 Flatfile was used. The RF model was optimized through Grid Search and 5-fold cross-validation, achieving high predictive performance with R² scores of 0.948 on training data and 0.928 on test data. Feature importance analysis indicated Mw and ClstD as the most influential parameters. The study demonstrates the robustness, accuracy, and generalization capability of the RF model, confirming its potential as a valuable tool in seismic risk analysis and providing a foundation for future development of more adaptable, data-driven models in earthquake engineering.

Received: 24 July 2025 | Revised: 25 November 2025 | Accepted: 5 December 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

Data are available from the corresponding author upon reasonable request.

Author Contribution Statement

Asadullah Ziar: Conceptualization, Methodology, Software, Formal analysis, Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization, Project administration. Ender Basari: Conceptualization, Validation, Writing - review & editing, Supervision.