Study On Investor’s Behavior Based On Computational Finance Experiments

As international financial integration, financial electronic and financial visualization, financial risk, especially financial crisis which erupted by unexpected serious financial events, will directly threaten financial security, economic security and even national security. Unfortunately, the traditional methods that based on standard financial theory can not effectively control the market risk, and the irrational behaviors of inves’tors in the market lead to the spread of crisis, or even market collapse.Although Behavior Finance, which had introduced the results and methods of cognitive psychology, can identify investors’ systematic cognitive biases and their behavior characterized, and point out that investor behavior is the most critical factor that leads to market price volatility, it can not correctly describe and track investor behavior and changes, and also can not systematically reveal the relationship mechanism between investor behavior and market price behavior. Agent-based Computational Finance, which based on complex adaptive theory and computational experiment method, can provide the necessary and complete supplements to Behavior Finance from theories and methods.This paper is a useful attempt to study the interaction between investor behavior and market price by using computational experiment method. From the complex adaptive theory of systematic science, we build an artificial financial market based on evolution of Agent’s behavioral heterogeneity using computer simulation and artificial intelligence technology from bottom to up, which embedded some cognitive patterns and herding behavior based on small-world network. In this computational experiment platform, we reveal the "emerging" of the complex features of financial market and its causes through repeatable controlled experiements, then provide some corresponding strategies of behavior control. Details are as follows:First, we provide the theoretical basis for study of Behavior Finance by using computational experiement method. This paper reviews standard finance theory and its limitations and discusses the framework construction of new finance theory and its status. Based on the theories and methods of Behavior Finance and Agent-based Computational Finance, this paper defines and characterizes the concept of investors’ bounded rationality and the nature of its evolution, introduces the modeling methods about Agent-based artificial financial market, and summarizes the calibration of artificial financial market and the advantages of computational experiment method. Subsequently, we construct a computational experiment platform that be used to study the relationship between investor behavior and market price behavior. Combined with the two branches of individual behavior modeling methods in Computational Finance, heterogeneous agent model and Santa Fe artificial stock market, this paper builds an artificial financial market based on evolution of Agent’s behavioral heterogeneity which embedded some cognitive patterns, and realizes Agent’s self-adaptive learning through evolving its prediction rules by using genetic algorithm and generation function. Simulating the artificial market, after utilizing market fraction analysis and market nonlinear characteristics tests successively, this paper calibrates the artificial market, and realizes Agent’s interactive and self-adaptive learning based on herding behavior, personal experience and market situations through embedding investor herding behavior based on small-world network, setting the relationship network between Agents and the contagion mechanism in herding behavior. By comparing the market price behavior, which before and after embedding herding behavior in the artificial market, to decompose the collaborative herding behavior, this paper reveals the impact of herding behavior on the market as well.Based on this controlled experiment platform, we design two-stage computational finance experiments about investor’s personal behavior and collaborative herding behavior. Through repeated experiments that independent of each other, this paper compares the complex characteristics of market price behavior that before and after adjusting parameters about individual behavior characteristics, individual behavior evolution and collaborative herding behavior, digs out some important behavioral order parameters, which is called as market sentiment, return-expected of market value, speed of market evolution, market information exchange mechanism, and investor type, investor attention and infected degree, then provides some corresponding strategies of behavior control after some parameters analyses. These strategies include behavior monitoring index construction and control strategy formulation, which help financial regulators, financial institutions and investors to understand the inner mechanism of the financial phenomenon and improve the capability about controlling financial risk.Finally, we summarise the significance about the study of Behavior Finance based on computational experiment method. Combined with the main research results and innovations, this paper points out two prospects for future research, which are improving the construction of artificial financial market and solving the "parallel implementation" problem.