Discussion Papers 2022

CIRJE-F-1207	"Multi-Agent Model Based Proactive Risk Management For Equity Investment"
Author Name	Mita, Daiya and Akihiko Takahashi
Date	December 2022
Full Paper
Remarks	Revised in June 2023; Forthcoming in Engineering Applications of Artificial Intelligence, Volume 125, October 2023.

Abstract
Developing and applying new artificial intelligence (AI) techniques in finance has become popular and one of the growing areas. Although many studies focus on return prediction and do not pay much attention to price formation, revealing its mechanism is essential in risk management, particularly in proactive risk management for investment to improve the performance. Thus, this paper introduces a novel multi-agent model, which is able to explain how agents’ portfolio rebalances determine the market price dynamics to clarify the price formation by applying a state space model. The technical novelty is the effective integration of state space modeling and fuzzy logic into a multi-agent model with four types of typical investors and their fuzzy trading strategies. By using the estimated unobservable fund flows of each trader in the model, this work proposes a new proactive warning signal. As a result, the signal improves both the risk and return of the investment in the Japanese and United States equity markets. Our findings indicate that the agents’ estimated fund flows driving asset prices help us to avoid a market crash, reduce the risk and improve the return in investment practice.
Keywords: multi-agent model, fuzzy system, state space model, proactive risk management, equity investment

Abstract

Developing and applying new artificial intelligence (AI) techniques in finance has become popular and one of the growing areas. Although many studies focus on return prediction and do not pay much attention to price formation, revealing its mechanism is essential in risk management, particularly in proactive risk management for investment to improve the performance. Thus, this paper introduces a novel multi-agent model, which is able to explain how agents’ portfolio rebalances determine the market price dynamics to clarify the price formation by applying a state space model.
The technical novelty is the effective integration of state space modeling and fuzzy logic into a multi-agent model with four types of typical investors and their fuzzy trading strategies. By using the estimated unobservable fund flows of each trader in the model, this work proposes a new proactive
warning signal. As a result, the signal improves both the risk and return of the investment in the Japanese and United States equity markets. Our findings indicate that the agents’ estimated fund flows driving asset prices help us to avoid a market crash, reduce the risk and improve the return in
investment practice.

Keywords: multi-agent model, fuzzy system, state space model, proactive risk management, equity investment