Publication

Abstract : This study investigates the application of deep learning techniques, specifically Convolutional Neural Networks (CNN) as well as Long Short-Term Memory (LSTM) networks, for earthquake prediction using a binary dataset. Predicting Earthquakes is a difficult task because of various factors the complex and non-linear nature of seismic data, and traditional models often fall short in capturing the intricate patterns involved. In this research, a 1D CNN is utilized to automatically extract spatial features from time-series seismic data, effectively reducing the need for manual feature extraction. The CNN is designed to detect local patterns in the data that might indicate upcoming seismic events. Meanwhile, the LSTM element is integrated to capture the temporal addictions as well as long-term trends extant in the seismic signal, given its ability to hold information over prolonged periods. The hybrid CNN-LSTM model aims to predict the occurrence of earthquakes by classifying each instance as either an imminent earthquake (positive class) or no earthquake (negative class) within a specific time frame. The binary nature of the dataset simplifies the prediction task, allowing the model to focus on learning the critical features associated with earthquake precursors. Preliminary results indicate that the integrated CNN-LSTM model is capable of identifying patterns in the data this paper obtained an accuracy of 97% which shows the feasibility of the proposed model.