Tokyo Metropolitan University
Tokyo Metropolitan University
National Astronomical Observatory of Japan
Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency (JAXA)(ISAS)
National Institute of Technology, Oita College
Shizuoka University
Tokyo Metropolitan University
出版者
宇宙航空研究開発機構(JAXA)
出版者（英）
Japan Aerospace Exploration Agency (JAXA)
雑誌名
宇宙航空研究開発機構研究開発報告: 宇宙科学情報解析論文誌: 第6号
雑誌名(英)
JAXA Research and Development Report: Journal of Space Science Informatics Japan: Volume 6
Deep moonquake occurs at depth of about 1000km, and it is most frequent lunar seismic event. Due to considerable noises and low amplitudes in deep moonquake waveforms, we have some difficulties to locate the sources when we use conventional method. In this paper, we have investigated suitable machine learning methods to classify the deep moonquake sources based on similarity among the waveforms. The machine learning-based method is more useful to reduce the computational time to classify the sources compared with the conventional method, and it has also advantage to apply the some types of effective features such as power spectral density for the classification. We compared performances of some machine learning methods in order to find the suitable method. Then, the deep events misclassified from some machine learning methods has been also investigated, and we have analyzed causes of misclassification (e.g., mislabel, outlier) for improvement of the classification. It can be expected that the suitable machine learning method will enable us to create a new moonquake event catalog and identify the unclassified events.
内容記述
形態: カラー図版あり
内容記述（英）
Physical characteristics: Original contains color illustrations
深発月震源分類に適した機械学習法の研究
AA1630049005.pdf
(1.2MB)
