An Artificial Intelligence based framework to predict Tsunami using Earthquake Characteristics

Abstract

Natural disasters like tsunamis create havoc among people and lead to several financial and human crises. This research aims to predict tsunami using Machine Learning and Deep Learning algorithms. We chose this work because early detection can save many lives. The dataset used in this work is obtained from the Kaggle website. The dataset includes earthquake details like magnitude, depth, latitude, longitude, CDI (Community Determined Intensity), MMI (Modified Mercalli Intensity), sig (significance), NST (number of stations), DMIN (distance to the nearest station) and gap along with the binary target variable to predict tsunami. We used several machine learning and deep learning algorithms for tsunami prediction and we made comparative study on the same. The algorithms used are: Logistic Regression, Decision Tree, K-Nearest Neighbors, Random Forest, Support Vector Machine, Feed forward Neural Network, Convolutional Neural Network and Recurrent Neural Network. It was observed that Random Forest outperformed all other algorithms with an accuracy of 0.9045, precision value of 0.8592, recall value of 0.9242 and F1-Score of 0.8905. It was observed that Decision tree performed well with an accuracy of 0.879, precision of 0.8507 and F1-Score of 0.875. Among Neural Networks, It was observed that the Feedforward Neural Network outperformed all other algorithms with an accuracy of 0.828, precision value of 0.7468, recall value of 0.8939 and F1-Score of 0.8137.