Statistical analysis of oscillation phenomenon in time series data

2023 Fiscal Year Final Research Report

• Project/Area Number: 19K20220
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 60030:Statistical science-related
• Research Institution: The University of Tokyo (2019, 2022-2023); Institute of Physical and Chemical Research (2020-2021)
• Principal Investigator: Matsuda Takeru 東京大学, 大学院情報理工学系研究科, 准教授 (50808475)
• Project Period (FY): 2019-04-01 – 2024-03-31
• Keywords: 時系列解析 / 状態空間モデル / 振動子

Outline of Final Research Achievements

We worked on the application and improvement of the oscillator decomposition, which uses a state-space model to extract oscillator components latent in time series data in a data-driven manner. In the application to infant fNIRS data, we obtained three types of oscillator decomposition: oscillators derived from brain activity, those corresponding to pulse waves, and those corresponding to mirrored noise. In the application to seismic wave data, oscillators with a period of about 11 seconds corresponding to volcanic long-period microtremor events were detected. We also ported the oscillator decomposition program from MATLAB to python and published the code on github.

Free Research Field

統計学

Academic Significance and Societal Importance of the Research Achievements

振動現象は自然界に遍在しており、その理解は分野を越えて重要である。本研究では、時系列データに潜む振動現象の理解・予測・制御を行うための統計手法を開発し、多様なデータに応用した。本手法によって、データ駆動的に振動現象の定量的理解が得られ、さらに不確実性を考慮した予測・制御を行うことが可能となる。