Ideal Agent System with Triplet States: Model Parameter Identification of Agent-Field Interaction

Area of Science:

• Econometrics
• Statistical Physics
• Computational Finance

Background:

• Company-specific news causes stock price movements independent of market trends, indicating sudden risk events.
• Statistical physics concepts are increasingly applied to econometrics for modeling socioeconomic phenomena.

Purpose of the Study:

• To extend existing agent-based models by including a "hold" position alongside "buy" and "sell" stances.
• To develop a theoretical framework for modeling agent behavior in capital markets using statistical ensembles.
• To enable accurate, one-step-ahead prediction of sudden risk events on capital markets.

Main Methods:

• Developed a theoretical model using microcanonical and canonical ensembles for an ideal agent system.
• Designed a procedure to estimate model parameters from capital market time series data.
• Applied the model to a capital market example for risk event assessment.

Main Results:

• The proposed model effectively incorporates "buy," "sell," and "hold" investor stances.
• A method for estimating model parameters from empirical data was successfully designed.
• The model demonstrated strong performance in one-step-ahead prediction of sudden risk events.

Conclusions:

• The developed model is well-specified for analyzing capital market dynamics.
• The model parameters are accurately determined, leading to reliable predictions.
• The approach offers a robust framework for assessing the impact of sudden risk events.