CapsNet and KNN-Based Earthquake Prediction Using Seismic and Wind Data

Abstract

Earthquakes are one of the worst natural hazards that frequently turn into disasters that cause extensive destruction and losses of human lives. Usually, sensor data are used for earthquake prediction. The system being developed would help predict an estimate of how the seismic activity is and what it would be after a specific period. In this chapter, the seismic data obtained from sensors, wind data, and longitude and latitude information are used to predict an earthquake in India. The longitude and latitude information is fed to the K-nearest neighbor (KNN) classifier to obtain one of the five seismic zones. The most suited and latest neural network (NN) termed capsule network (CapsNet) is then used for prediction. CapsNet accepts the seismic and wind data for intermediate earthquake prediction. Eventually, it uses the blender network that combines the output of KNN and CapsNet for final earthquake prediction. The proposed method is compared with present cutting-edge earthquake prediction systems concerning performance and results. The comparison reveals that the proposed approach obtains promising and reliable results. In many parameters, it outperforms the other earthquake prediction systems.