Prediction of Tsunami Event at Early Stage Using Ensemble Bagging Random Forest Classifier

Earthquakes and tsunamis are among the most destructive natural disasters. Early detection is essential to mitigate loss of life and property damage. This research work mainly focuses on the ways to improve the accuracy of early detection of the Tsunamigenic potential of earthquakes by analyzing the earthquake seismogram signal. This research could lead to more effective and timely warnings, potentially saving lives and reducing damage caused by tsunamis. The seismic features are extracted from the primary wave (P-wave) of the signal and fed into the classifier. The characteristics of each feature are studied to know whether the individual selected features can differentiate the tsunami signals from the non-tsunami signals. The ensemble bagging random forest classifier has been used to classify the P-wave into tsunami and non-tsunami. The accuracy of the prediction of non-tsunami waves is about 92%, but the overall accuracy of the classifier is 96.62%. The classifier predicts the tsunami waves with better accuracy at about 99.6%. Moreover, it accurately forecasts the tsunami waves within 2.75[Formula: see text]s after the earthquake occurs. This suggests that the proposed technique is suitable for forecasting a tsunami soon after an earthquake occurs.