Comparative Analysis of Machine Learning Models for Earthquake Prediction Using Large Textual Datasets

Earthquake is one of the most devastating natural calamities known to man. Earthquakes can affect lives in unimaginable ways and predicting them well in time is one of the most important things when it comes to earthquake damage reduction. Many approaches are used to predict earthquakes, and machine learning can aid in early and timely prediction of earthquakes and thus reducing any damage. This paper provides a comprehensive and comparative analysis of various classical machine learning algorithms in the prediction of earthquakes using a large textual dataset that holds information about all the historical earthquakes. This dataset holds many vital data and serves as the basis for training and evaluating various machine learning models. Different regression models including a random forest regressor, decision tree regressor, linear regressor, and a regression artificial neural network (ANN) were used for earthquake prediction. Machine learning architectures help in capturing the relationship between the independent and dependent variables, whereas the ANN captures more complex patterns in the data. Unlike linear regressor, decision tree and random forest regressors capture non-linear relationships between dependent and independent variables. A substantial amount of experimentation and evaluation of the models were conducted to compare and contrast the model performance based on relevant performance metrics. This research offers insights into application of several machine learning models in earthquake prediction and their real-world applicability.