AUTO CLASSIFICATION OF VOLCANO-TECTONIC (VT) EARTHQUAKE USING MACHINE LEARNING CASE STUDY OF AGUNG VOLCANO UNREST 2017

• Main Author: Martanto
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/62053
• Institution: Institut Teknologi Bandung

Volcano is a complex system and to understand its characteristics it is necessary to carry out activities such as research, investigation, and monitoring. This whole series of activities is carried out with the aim of reducing the risk of volcanic hazards. Of the three activities, volcano observation is a routine activity that is carried out 24 hours every day and involves many methods, such as sesimic, magnetic, deformation, geochemical, thermal, and or a combination of these methods. Of all these methods, the seismik method is used as the main method in determining the level of activity. This method is used because it can provide real time information, can represent the subsurface conditions of the volcano, and can be used to calculate the amount of energy contained in the volcano. This energy representation is represented in various types of earthquakes, one of which is a volcanic earthquake (Volcano-Tectonic) or commonly abbreviated as VT. The more VT earthquakes, the higher the level of volcanic activity. Counting the number and identification of VT earthquakes has been done manually by reading analog and digital seismogram data. The problem arises when a volcano is experiencing a unrest phase where the recorded earthquakes can number in the thousands and identification of VT earthquakes takes a very long time. Meanwhile, a risk analysis based on these data is needed as quickly as possible. This research uses VT earthquakes from the Mount Agung unrest (1 September - 31 December 2017) by applying the machine learning (ML) using Random Forest Classifier method to help auto-classify VT earthquakes. Using RFC, the prediction accuracy for VT and non-VT earthquake classifications reached 99.53% with the smallest RMSE of 0.068734.