A multi-agent based simulated stock market - testing on different types of stocks

Abstract

Previously, we have developed a multiagent based simulated stock market where artificial stock traders coevolve by means of individual and social learning and learn to trade stock profitably. We tested our model on a single stock (British Petroleum) from the LSE (London Stock Exchange) where our artificial agents demonstrated dynamic learning behaviours and strong learning abilities. We extend our previous work by testing the model on different types of stocks from different sections of the stock market. The results from the experiments show that the artificial traders demonstrate stable and satisfactory learning abilities during the simulation regardless of the different types of stocks. The results lays the foundation for our future work - developing an efficient portfolio manager from a multiagent based simulated stock market.