A Comparative Study of Machine Learning Models for Predicting Earthquake Magnitudes

Title

A Comparative Study of Machine Learning Models for Predicting Earthquake Magnitudes

Creator

Shaw, Ayush; Shoran, Preety; Vidushi

Description

This study evaluates the effectiveness of machine learning models in predicting earthquake magnitudes, aiming to address the challenges posed by traditional seismological methods. By leveraging geospatial and temporal data, the research compares the performance of Random Forest (RF), Artificial Neural Network (ANN), and K-Nearest Neighbours (KNN) using Mean Absolute Error (MAE) as the primary metric. Random Forest demonstrated the lowest MAE of 0.3291, showcasing its ability to handle complex, non-linear patterns better than ANN (0.3660) and KNN (0.3758). The analysis highlights the importance of geospatial and temporal factors in improving prediction accuracy, offering insights into their predictive significance. While traditional methods struggle with high-dimensional data, this study eliminates these limitations by employing machine learning models capable of extracting meaningful patterns. These findings underscore the potential of ensemble methods like Random Forest for enhancing earthquake prediction systems. Future research will explore hybrid approaches and real-time data integration to further advance predictive accuracy in seismology. 2025 IEEE.

Source

International Conference on Engineering, Technology and Management, ICETM 2025;

Date

01-01-2025

Publisher

Institute of Electrical and Electronics Engineers Inc.

Subject

Earthquake Prediction; K-Nearest Neighbours; Machine Learning; Magnitude Prediction; Temporal and Geospatial Data

Coverage

Shaw A., Christ (Deemed to be University), Bangalore, Department of Computer Sciences, Delhi, India; Shoran P., Christ (Deemed to be University), Bangalore, Department of Computer Sciences, Delhi, India; Vidushi, Christ (Deemed to be University), Bangalore, Department of Computer Sciences, Delhi, India

Rights

Restricted Access; Hardcopy may be available in the library

Relation

ISBN: 979-833156668-5;

Format

online

Language

English

Type

Conference paper

Identifier

https://doi.org/10.1109/ICETM63734.2025.11051600

https://www.scopus.com/pages/publications/105012020590?origin=resultslist