Classifying volcanic earthquake types of Mount Merapi using Artificial Neural Network (ANN)

Abstract

Indonesia, a country that passed by ring of fire, make this country have a lot of volcanoes. Therefore, monitoring volcanic activity is needed to dig deeper into volcano-science and to help in mitigating the risk of volcanic disasters that can be done by utilizing today’s technological developments. By utilizing of artificial inteligence and machine learning technology that can automate the processing of large amount of data, monitoring of volcanic activity can be made easier. On this research, the use of artificial neural network (ANN) for classifying volcanic earthquake patterns from Mt. Merapi. Some volcanic events from Mt. Merapi was be categorized into five classes: volcano-tectonic A, volcano-tectonic B, multiphase, low frequency and rockfall. The seismic data record from Mt. Merapi was processed becoming frequency spectrum as feature space and input for the artificial neural network. The output of the artificial neural network is one type of Mt. Merapi seismic event. The input data consist of 430 volcanic events record of Mt. Merapi during October 2019. Some architecture model of artificial neural network has been developed too. The experiment result of the artificial neural network for these volcanic events with ratio training and testing around 70%:30% shows accuracy was >83.7%. This accuracy is higher than previous studies of machine learning using same data with XGBoost algorithm with accuracy 76.7% and SVM algorithm with accuracy 74.4%.