Harvesting random embedding for high-frequency change-point detection in temporal complex systems.

Area of Science:

• Complex systems analysis
• Time-series data analysis
• Network science

Background:

• System dynamics are heavily influenced by temporal structures.
• Accurate detection of structural changes is vital for understanding, modeling, and predicting evolving systems.
• A model-free approach is needed for real-world applications using only time-series data.

Purpose of the Study:

• To develop a practical, model-free method for detecting temporal change points in complex systems.
• To address the challenge of identifying system structural changes without prior knowledge of system equations.

Main Methods:

• Developed a novel model-free approach named temporal change-point detection (TCD).
• Integrated dynamical and statistical methods.
• Exploited spatial information from high-dimensional time-series data.

Main Results:

• The TCD approach successfully detects separate change points in systems without needing a priori information.
• TCD can identify high-frequency emergent change points, surpassing existing methods.
• Effectiveness demonstrated across diverse real-world systems (biology, geology, social science).

Conclusions:

• The proposed TCD method offers a practical and effective solution for model-free temporal change-point detection.
• TCD enhances the ability to analyze and predict complex system dynamics by accurately identifying structural shifts.
• This approach has broad applicability in various scientific domains.