Bayesian Calibration of Hyperparameters in Agent-Based Stock Market

In this paper, we constructed an agent-based simulation of the stock market where traders are affected by neighbors' decision in different size groups and public information. In order to reproduce the stylized facts of the real market, agent-based model require a calibrated set of parameters which can regenerate the real data with simulation models. Bayesian optimization is introduced to tune the hyperparameters of the traders' behavior as well as of the environment. The experimental results on Bayesian calibration demonstrated that the proposed separate calibrations reduce simulation error, with plausible estimated parameters. With empirical data of the Dow Jones Industrial Average index, statistical analysis of the simulated data showed that the model was able to replicate some of the important stylized facts in real markets such as: random walk price dynamics, the leptokurtosis, heavy tail of the returns and volatility clustering.