Major Earthquake Event Prediction Using Various Machine Learning Techniques

According to the extent or time period taken to foretell the occurrence of an earthquake, two primary classifications of methods are used. In other words, short-term maps are generated over the course of the few hours or, at most, days while maps that are made months or even years in advance are referred to as forecasts. The majority of research is aimed at the predictions that take into account the history of the earthquake in the given regions or countries. The objective of the present investigation is to identify if an earthquake event is positive or negative using different algorithms. The implementation of Random Forest, Naïve Bayes, Logistic Regression, Multi-Layer Perceptron, AdaBoost, K-Nearest Neighbors, SVC using Support Vector Machine, and Classification and Regression Trees has been performed to a real earthquake data set. Different hyperparameters for each algorithm were selected and the accuracy of each of them was compared based on several measures. The features were identified to have average values of 0.89, 0.89 and 0.83 for significant earthquake activities by three of the models. Some of the other algorithms that have been used in conjunction with the proposed approach are Random Forest, Naïve Bayes, Logistic Regression, AdaBoost, KNN, Support Vector Machine and Multi-Layer Perceptron Classifier.