Machine Learning-Based Seismic Activity Prediction

Abstract

Earthquakes can have devastating consequences, causing ground shaking, landslides, and changes in landscapes. Fault line ruptures can alter river courses and disrupt infrastructure, while underwater earthquakes may trigger tsunamis, affecting coastal ecosystems and communities. Liquefaction can temporarily weaken the ground, leading to structural damage, and aftershocks can further exacerbate existing damage and hinder recovery efforts. Human impacts are significant and can result in injuries, fatalities, displacement, and psychological trauma. Economic consequences can involve disruption to industries and livelihoods, while response and recovery efforts may have environmental consequences. This chapter focuses on earthquake prediction using various parameters such as date, time, latitude, longitude, depth, and magnitude. The authors have used a world map as a dataset to train our model, where we predict earthquakes using gradient boosting regressor. They have broken down the complex and challenging problem into simpler like mean squared error (MSE) as a loss function, accuracy, precision, recall, F1 score, confusion matrix. As of our last knowledge update in September 2023, earthquake prediction remains a field of ongoing research and does not have precise predictive models. The advantage of this model is its accuracy which is predicted as output. However, by comparing actual datasets with predicted outcomes of occurrences, we can identify risk free areas for livelihood. The proposed model achieved an accuracy of 86.1% and 99.7% in terms of magnitude and depth, which is higher than the accuracy of existing earthquake prediction methods.