Uji Performa Prediksi Gempa Bumi di Jawa Timur dengan Artificial Neural Network

Abstract

East Java Province is an area directly adjacent to the Eurasian and Indo-Australian plate subduction zones, this has resulted in East Java province being an area prone to earthquakes. Predictions regarding the frequency of earthquake occurrences are very interesting to study. This needs to be done in order to increase our preparedness in an effort to reduce the risk of earthquakes. Research on earthquake prediction has been carried out, one of which is the artificial neural network method. The purpose of this study is to obtain the best network architecture that is applied to the data on the frequency of earthquake occurrences per month in East Java Province. Data on earthquake occurrences come from the BMKG Nganjuk Geophysics Station, which was recorded during the 2016-2021 period. The data was then grouped into the total frequency of events per month. The criteria for selecting the best network architecture are carried out by comparing each possible architecture's error values. The test method uses SSE (sum square error) criteria for each architectural model of the artificial neural network. The test results show that the input variation has a significant contribution and a greater correlation than the variation in the number of hidden neurons. The best prediction results are obtained in the model with 9-30-1 architecture with an error value of 0.1958.