Earthquake Prediction using Supervised Machine Learning Techniques

Abstract

Earthquakes are one of the greatest natural catastrophes which cause a gigantic loss of human
life and property. Earthquake Forecasting is an extremely essential complication in seismology,
the realization of which will undoubtedly save abounding human lives from seismology, the
accomplishment of which can assuredly protect countless human lives. Various solutions,
including machine learning algorithms, deep learning algorithms, and mathematical analysis,
have been offered to deal with this issue. Unfortunately, machine learning techniques lack
satisfactory outcomes as an aftermath of the apparently versatile and erratic calamity of tremor.
As a result, it is important to consider historical data from a wider area while conducting tremor
prediction for a specific locality. This research used a collection of previous earthquakes to
predict patterns and future earthquakes. Machine Learning algorithms are used to simulate the
order of earthquakes. Due to the complexity of predicting the future earthquake, the aim is to
develop and analyze the machine learning algorithm, methods, framework, and tools used to
predict the earthquake using different parameters. In this work, powerful Machine learning
techniques along with the data science process are used to predict the seismic events exactly.
A classification mechanism of machine learning techniques namely Logistic Regression,
Random Forest algorithm, Naïve Bayes classification and decision trees are separately applied
to the real earthquake dataset, to pick out the best model. A number of hyperparameters have
been chosen for each training model, and the acquired prediction results have been fairly
compared and evaluated using several metrics, resulting in a trustworthy forecast. Furthermore,
in order to get the outcome, factors including accuracy, precision, recall, and f1 score of
confusion metrics and performance measures are assessed. The above methods for predicted
earthquakes yield significant and encouraging results with the best accuracy than the ensemble
methods used in the previous work for predicting the seismic, and signaling advancement
towards the ultimate robust prognostication process.