IIIT Hyderabad Publications |
|||||||||
|
FORECASTING OF EARTHQUAKE USING RECURRENT NEURAL NETWORKAuthors: Swapnil Nayan, Prakke Bharath,Aniket Bhalkikar,Pradeep Kumar Ramancharla Conference: 17 th World Conference on Earthquake Engineering (WCEE 2020) Pages: 1-9 Location Japan Date: 2020-09-13 Report no: IIIT/TR/2020/88 AbstractEstimating the time, location and the size of an earthquake event using neural networks is an emerging technology. Present study aims to predict the near future earthquake events in each of the regions partitioned as clusters of epi-central locations of earthquakes related to Indian subcontinent. K-means clustering algorithm is applied to develop a catalogue of earthquake epicenters. Two approaches are considered for the purpose of forecasting. In the first approach regions obtained as clusters are considered as the area sources. The distribution of time series of (temporal distribution) events are divided into five time periods viz., 15 days, 30 days, 60 days, 90 days and 180 days. The accuracy of forecasting a magnitude for the next time period is studied using application of Recurrent Neural Network In the second approach the seismic data of each cluster is further grouped into three categories of magnitude viz., minor (≤ 4 Mw), moderate (between 4 Mw and 6 Mw) and strong (> 6 Mw). Accuracy of forecasting for each of the group belonging to the clusters is computed considering the same five time periods. The epicentral locations represented as latitude and longitude is the inputs for recurrent neural network. The results show that accuracy of second approach is significantly more than first approach. Finally, an attempt has been made to compare the accuracy level of Recurrent Neural Network method with the accuracy level of Gutenberg Richter law in predicting the earthquakes for future 60 days. It has been observed that Gutenberg-Richter law poorly forecasts near future earthquake. Keywords: Recurrent Neural Network; Earthquake forecasting, Clustering Full paper: pdf Centre for Earthquake Engineering |
||||||||
Copyright © 2009 - IIIT Hyderabad. All Rights Reserved. |