Spatio-Temporal Seismic Data Analysis for Predicting Earthquake: Bangladesh Perspective

Abstract

Earthquake prediction concerns specifying the earthquake's occurrence time, location, latitude, longitude, and intensity level. The determination of factors for the next earthquake happening in a region is very hard, almost impossible because earthquake occurrence depends on many things, such as changes in global warming, underground seismic wave, underground explosions, and underground rocks colliding, etc. But, nowadays, many types of research have been done around the world to build an earthquake warning system which upon detection of an earthquake, provides a real-time warning to the neighboring regions that might be affected. In this study, only the Spatio-temporal seismic data of Bangladesh is analyzed to propose an earthquake prediction model using the probabilistic assumption of the next earthquake happening in and around the Bangladesh region. The experimental dataset contains 100 years of a historical earthquake happening records in and around Bangladesh from the year 1918 to 2018 and this data is collected from Bangladesh Meteorological Department's (BMD) climate division. A comparative and comprehensive analysis is done to identify the best-suited model for earthquake prediction using some pioneer computationally intelligent and probabilistic machine learning algorithms, such as support vector machine, random forest, decision tree, naïve Bayes, and k-nearest neighbor.