Application of EfficientNet Architecture for Geophysical Event Classification Using Transfer Learning Convolutional Neural Network Method
Abstrak
Volcanic activity can be identified using volcanic earthquake events. Volcanic earthquakes generally have similar characteristics; a method of classifying seismic events is required. This research explores the practical application of a geophysical event classification system that leverages the EfficientNet-B0 architecture within a Transfer Learning Convolutional Neural Network framework. The classification method focuses on categorizing events such as earthquakes, rockfalls, and gas bursts from spectrogram images, aiming to enhance accuracy and efficiency. The study utilizes a dataset of 5,288 geophysical event records sourced from the Institute for the Center for Research and Development of Geological Disaster Technology in Indonesia (BPPTKG). This research utilizes several parameters to enhance the accuracy and efficiency of the models. These parameters include batch size, image size, dropout rate, activation function, learning rate, optimization type, and number of epochs. The results showed that the CNN model with the EfficientNet-B0 architecture and without transfer learning achieved an accuracy of 92%. Meanwhile, the CNN model using transfer learning and the same architecture attained an accuracy of 89%. A key contribution of this research is its innovative use of EfficientNet-B0, which optimizes computational speed and accuracy in handling geophysical event classifications. This research provides valuable insights for the development of automated disaster monitoring systems, which can significantly enhance disaster mitigation strategies by offering real-time event classification. The classification results can be applied to the timely prediction of natural disasters, thus contributing to more effective response and preparedness efforts in disaster-prone regions.
Identifier
- OpenAlex: https://openalex.org/W7154512328
- DOI: https://doi.org/10.62527/joiv.10.2.3937