Bayesian Inference and Logistic Regression Based Modelling for Earthquake Probability Estimation in East Java

Abstrak

East Java is one of the seismically active regions in Indonesia, yet predictive studies that integrate spatial data and event parameters remain limited. This study develops a two-stage approach to model earthquake risk more comprehensively by combining Bayesian inference and logistic regression. The first stage employs a Bayesian model to estimate the daily probability of earthquake occurrence based on historical data from 2014 to 2024. The results show an average daily probability of 13.5%, with a 95% credible interval indicating a high level of confidence. Spatially, Region 1 (covering southern East Java) is identified as the area with the highest probability, followed by Region 3 and Region 2. In the second stage, logistic regression is used to identify combinations of event parameters—particularly magnitude and depth—that significantly influence the likelihood of moderate-to-major earthquakes (magnitude ≥ 5.0). The prediction results indicate that most high-risk events occur at shallow depths in Region 1 and Region 3, while Region 2 appears less frequently but still presents underlying geological hazards. These findings demonstrate that integrating probabilistic modeling with parameter-based classification offers a more refined understanding of earthquake risk. As an initial framework, this study also opens avenues for developing future early warning systems based on dynamic data and machine learning methods.

Identifier

• OpenAlex: https://openalex.org/W4415649304
• DOI: https://doi.org/10.20885/eksakta.vol6.iss2.art4