Experimental analysis of earthquake prediction using machine learning classifiers, curve fitting, and neural modeling

Abstract

Abstract An earthquake is one of the most massive natural disasters which happens unexpectedly shaking the earth's surface. Due to earthquakes, not only infrastructure but also buildings get damaged thereby affecting lifestyle. For the early-stage prediction of the earthquake impact, machine learning can play a vital role, and this entails the novelty of the work. For this perception, six different machine learning classifiers namely Artificial Neural Network, Random Tree, CHAID, Discriminant, XGBoost Tree, and Tree-AS on six datasets of different regions of India. All the algorithms have been applied to each dataset. The objective of the research is to predict the value of magnitude for the future earthquake in India and nearby regions from the historical data on earthquakes. From the result, It has been observed that for Andaman & Nikobar dataset XGBoost Tree achieved the highest accuracy with 99.10%, for the Gujarat dataset Tree-AS achieved the highest accuracy with 91.67%, for the North India dataset Artificial Neural Network achieved the highest accuracy with 99.13%, for North East India dataset XGBoost Tree achieved the highest accuracy with 99.04%, for Nepal-UP-Bihar dataset XGBoost Tree achieved the highest accuracy with 99.01%, for Nearby India’s Country dataset XGBoost Tree achieved the highest accuracy with 92.12%. From all the results, it has been noted that XGBoost tree classifier performed well in most datasets., the Curve has been made between magnitude & gap, magnitude & magnitude error, and magnitude and depth error for finding the mathematical relation between them.