Zero-Inflated Integer Autoregressive (ZINAR) Approach in Modeling Major Earthquakes in Sumatra Region

Sumatra is one of the regions in Indonesia with the highest seismic activity due to the convergence of the Indo-Australian and Eurasian plates. Major earthquakes with magnitudes of more than 7 pose serious risks, making it essential to understand their frequency for effective disaster mitigation planning. This study aims to model the frequency of major earthquakes in Sumatra using the Zero-Inflated Integer Autoregressive (ZINAR) model, which accommodates discrete data with excess zeros and temporal dependence. The analysis indicates significant overdispersion and zero inflation, leading to the selection of the ZINAR model as the most suitable approach compared to alternative models such as INAR, NB-INAR, and NB-ZINAR. The ZINAR model demonstrates superior performance in capturing the occurrence patterns of large earthquakes and effectively identifies non-event periods, although its predictive accuracy for actual event timing remains limited. Overall, the ZINAR model proves effective for modeling complex earthquake data and provides valuable insights to support disaster mitigation efforts in the Sumatra region.