A dual approach to earthquake STEAD dataset analysis: Regression and classification

P. Devasudha¹, R. Raghupathy², M. Vadivukarassi³

---

ABSTRACT

This research analyzes the Standard Earthquake Dataset (STEAD) to predict earthquake magnitudes and severity. The research adopts a dual approach, utilizing both regression and classification methodologies. For regression, Random Forest Regression, Gradient Boosting Regression, Linear Regression, and Support Vector Regression models are utilized. Key features extracted from STEAD include 'receiver latitude', 'receiver longitude', 'lat_long_interaction', 'lat_squared', 'long_squared', 'lat_cubed', 'long_cubed', 'exp_lat', 'exp_long', and 'source_magnitude'. Gradient Boosting Regression demonstrates superior performance. For classification, the events are categorized into two classes: 'medium' and 'severe' based on source magnitude. This study employs Random Forest Classifier, Gradient Boosting Classifier, Logistic Regression, and Support Vector Machine (SVM) Classifier. The Random Forest Classifier exhibits high efficacy in distinguishing between mild and severe events. This research demonstrates the utility of Machine Learning (ML) in earthquake prediction, with Gradient Boosting Regression and Random Forest Classifier emerging as the most effective models. The performance of these models is evaluated using assessment metrics such as accuracy, precision, recall, and the F1-score.

Keywords: Regression, Classification, Random Forest, Classifier STEAD.

How to cite this article: Devasudha P, Raghupathy R, Vadivukarassi M. A dual approach to earthquake STEAD dataset analysis: Regression and classification. Int J Drug Deliv Technol. 2026;16(8s): 29-34; DOI: 10.25258/ijddt.16.8s.6

Source of support: None.

Conflict of interest: None