Asset Pricing under Evolution of Agent's Behavioral Heterogeneity in an Artificial Financial Market

Mi Yang,Chao-Qun Ma,Lin Zou
DOI: https://doi.org/10.1109/iciecs.2010.5678353
2010-01-01
Abstract:We use the study method of Computational Finance to explore the formation and evolution of asset prices from the standpoint of the evolution of investor individual's heterogeneous behavior through building an agent-based artificial financial market. In our model,agent will consider fundamental information and price tendency simultaneously at each period to form expectation that based on personal characters, such as mood,memory length,adjustment and extrapolation speed.The weight that he relies on both fundamental and technical analysis varies over time,which is the best prediction to current market state from empirical rule-set that has been updated through learning from the market situation with Genetic Algorithm (GA) and individual's trading experience with Generation Function (GF).The adaptive updating of the weight represents the evolution of agent's behavior.The model captures the two prime behaviors of agent and the trade-off between them,which realized by agent's adaptively personal learning. Simulation testing shows that even considering agent's variation of behavior in the market,the market fraction also has to be composed of the proportions of confident fundamentalists,chartists and adaptively rational agents as empirical evidence suggests, which will cause the so-called "stylized facts" in financial time series,under a market maker scenario.The impact of the market fraction varies on asset pricing also has been examined.
What problem does this paper attempt to address?