Intelligent agent-assisted adaptive order simulation system in the artificial stock market

Abstract

Agent-based computational economics (ACE) has received increased attention and importance over recent years. Some researchers have attempted to develop an agent-based model of the stock market to investigate the behavior of investors and provide decision support for innovation of trading mechanisms. However, challenges remain regarding the design and implementation of such a model, due to the complexity of investors, financial information, policies, and so on. This paper will describe a novel architecture to model the stock market by utilizing stock agent, finance agent and investor agent. Each type of investor agent has a different investment strategy and learning method. A prototype system for supporting stock market simulation and evolution is also presented to demonstrate the practicality and feasibility of the proposed intelligent agent-based artificial stock market system architecture.

Introduction

The financial turmoil triggered by the US subprime mortgage crisis has swept the world since 2007. Many banks, real estate investment trusts (REIT) and hedge funds have suffered significant losses as a result of mortgage payment defaults or mortgage asset devaluation. Some even collapsed, such as Bear Stearns and Lehman Brothers (Sorkin, 2008, White and Anderson, 2008). Jan Hatzius estimates that in the past year, financial institutions around the world have already written down $408 billion worth of assets and raised $367 billion worth of capital (Hilsenrath, Ng, & Paletta, 2008). The crisis has severely shaken people’s faith in traditional economic theory. “We have had a massive failure of the dominant economic model”, Eric Weinstein said. In 2009, Nature journal published two articles on agent-based modeling to study the economics and prevent the financial crisis (Buchanan, 2009, Farmer and Foley, 2009). Agent-based computational economics (ACE) is the computational study of economies modeled as evolving systems of autonomous interacting agents (Wu, 2001). ACE is a bottom-up culture dish approach to the study of economic systems (Tesfatsion, 2011). It has been applied to research areas such as asset pricing, stock market simulation, industry dynamics, and macroeconomics.

China’s economy has developed rapidly in the past 30 years. The healthy development of the stock market is very important for the national economy. However, changes in the stock trading mechanism may have a greater impact on the market. Thus, the Shanghai Stock Exchange launched China’s first innovation R&D and experimental platform based on finance simulation and modeling technology in 2011. The short-term goal of the innovation experimental platform is to construct a table-top exercises environment for business innovation research. The long-term goal is to build an open and professional R&D experimental environment that can provide support and service for continuous trading mechanism innovation. To achieve this goal, we need to build an innovation experimental platform by designing an adaptive simulation system based on intelligent agents. In practice, some researchers have already developed agent-based simulation systems of the stock market in past years (LeBaron, 2002, Nadeau, 2009, Wang et al., 2004). However, there are still two limitations in today’s practical simulation systems that need to be addressed. These are:

• The simulation of market news in the investment decision-making process. In general, investors make investment decisions through comprehensive analysis of various information in which the financial magazine is an important information source. However, little work has been done about the utilization of such information in the decision process of fundamentals investors. Thus, how to simulate the decision-making process of fundamentals investors based on financial information is one major challenge.

• The learning mechanism of fundamentals investors. In the long practice of investing, each investor will continue to learn to improve their profitability. Investors improve and optimize their strategies based on investment return. Everyone is willing to believe information sources that have strong predictive power. Thus, the predictive ability evaluation of the various information sources is a key problem. How to design and implement the learning mechanism of fundamentals investors is another challenge.

To address these challenges, we studied the conceptual model of the stock market in-depth. It includes stocks, investors, financial information, trading mechanisms and other participants. We use a different agent to represent each type of participant. The relationships among the agents are embodied using the agent hierarchy. We describe in detail the design of the stock agent, investor and financial agent, which shows how market news is used in the decision-making process. We study the investment strategy and learning algorithm of fundamentals investors and other types of investors. Finally, we design and implement one system to simulate the real stock market, i.e., Intelligent Agent-assisted Order Simulation System (IAOSS), and evaluate the reasonableness of the system design through some technical indicators.

The rest of the paper is organized as follows. Section 2 discusses related work. Section 3 describes the system design, including the conceptual model, agent hierarchy, agent design, and system architecture. Section 4 describes system implementation, including agent implementation and the investor learning algorithm. Section 5 describes the application and evaluation of the system and conducts short selling experiments. Finally, we draw conclusions to end this paper.