Tsunami Early Warning Detection using Bayesian Classifier

Tsunami is a term for disaster where the seawater rises to the land caused by the great earthquake with a shallow epicenter in the ocean. Indonesia is vulnerable to the tsunami; especially in the area where the junction of Eurasia, Indo-Australia, and Pacific Plates meet. Soon after the earthquake happened, there is a time interval before a tsunami to come. The break-time can be used to alert people to evacuate themselves from the Tsunami. This study proposed a tsunami early warning system, which autonomously predicts the potential of tsunami hazard using machine learning techniques. Bayesian classifier was trained to predict the tsunami potential. The tsunami training data was taken from the InaTews website; a national project of Indonesia that involves many institutions, local or international. InaTews informed the real-time data of earthquakes and tsunami that happened in Indonesia. The proposed methods extracted some invariant characteristics from the data and trained the machine learning classifiers to predict the potential result; either tsunami or not-tsunami using three parameters of earthquake: magnitude, epicenter, and location. The experiment gave an optimistic result; using a varying number of training data, it gained 92.37% on the average accuracy rate and 0.98 on the F1 score.