[P-EM19] Dynamics of the Inner Magnetospheric System

[PEM19-P17] Machine learning classification of low frequency waves observed by Arase satellite

*天野 駿¹、三宅 壮聡¹、笠原 禎也² (1.富山県立大学、2.金沢大学)

キーワード：磁気圏、科学衛星あらせ、機械学習

Various types of low frequency waves are observed by Electric Field Detector (EFD) onboard Arase satellite. In this study, we are going to detect these low frequency waves from EFD data, and classify these waves into several types by using machine learning.Various types of low frequency waves are observed by Electric Field Detector (EFD) onboard Arase satellite. In this study, we are going to detect these low frequency waves from EFD data, and classify these waves into several types by using machine learning.
At first, we use SVM method to detect low frequency waves from 24 hour plots of EFD spectrum data. We tested several parameters of SVM, and detected low frequency waves with more than 96% accuracy ratio.
We applied this SVM method to EFD spectrum data from 2017 to 2019, and succeeded in detecting 806 low frequency waves.
Next, we try to classify these detected low frequency waves into several types by clustering. We apply K-means method and hierarchical clustering method to the image data of EFD spectrum plot and the numerical data which are the duration time and the center frequency of the low frequency waves, respectively. Therefore, we found 5 types of low frequency waves with different characteristics.