Multi-Agent Forex Trading System
Automated trading is a novel field of study in which computer programs are put in charge of deciding when and how to trade financial instruments. Intelligent agents, with their ability to act autonomously and to adapt and interact with the environment, seem like an obvious choice for the development of automated trading systems. The aim of this article is to analyze how well intelligent agents suit this task. We implemented a set of autonomous currency trading agents, using an architecture that consists of an ensemble of classification and regression models, a case-based reasoning system and an expert system. A total of six trading agents were implemented, each being responsible for trading one of the following currency pair in the Forex market: EUR/USD, EUR/JPY, EUR/CHF, USD/JPY, USD/CHF and CHF/JPY. The agents simulated trades over a period of 23 months, having all achieved a reasonable profit trading independently. However, their strategies resulted in relatively high drawdows. In order to decrease the risk inherent to these high drawdowns, the same simulation was performed while making the agents share the monetary resources. As expected, this strategy of investment diversification originated better results. Still, when the trading costs were taken into consideration, the overall trading performance was less than impressive. That was due to the fact that each agent performed too many trades, and the cost associated with the trading commissions became prohibitively high. We were able to lessen the impact of the trading costs in the total profit by integrating the agents in a multi-agent system, in which the agents communicated with each other before opening new trades. This allowed them to calculate the intended exposure to the market, which in turn enabled them to avoid redundant trades. Under simulation and using low leverage, this multi-agent system obtained a 55.7% profit in 23 months of trading, with a 9.0% maximum drawdown.
KeywordsMove Average Trading Cost Financial Instrument Short Sell Knowledge Module
Unable to display preview. Download preview PDF.
- 2.Kimoto, T., Asakawa, K., Yoda, M., Takeoka, M.: Stock market prediction system with modular neural networks. In: 1990 International Joint Conference on Neural Networks, vol. 1, pp. 1–6 (1990)Google Scholar
- 3.Kwon, Y., Moon, B.: Daily Stock Prediction Using Neuro-genetic Hybrids. In: Cantú-Paz, E., Foster, J.A., Deb, K., Davis, L., Roy, R., O’Reilly, U.-M., Beyer, H.-G., Kendall, G., Wilson, S.W., Harman, M., Wegener, J., Dasgupta, D., Potter, M.A., Schultz, A., Dowsland, K.A., Jonoska, N., Miller, J., Standish, R.K. (eds.) GECCO 2003. LNCS, vol. 2723, pp. 2203–2214. Springer, Heidelberg (2003)CrossRefGoogle Scholar
- 4.Franses, P., Griensven, K.: Forecasting exchange rates using neural networks for technical trading rules. Studies in Nonlinear Dynamics and Econometrics 2(4), 109–114 (1998)Google Scholar
- 5.Lu, H., Han, J., Feng, L.: Stock movement prediction and N-dimensional inter-transaction association rules. In: 1998 ACM SIGMOD Workshop on Research Issues on Data Mining and Knowledge Discovery, pp. 1–7 (1998)Google Scholar
- 7.Tay, F., Cao, L.: Application of support vector machines in financial time series forecasting. International Journal of Management Science 29(4), 309–317 (2001)Google Scholar
- 8.Abraham, A.: Analysis of hybrid soft and hard computing techniques for Forex monitoring systems. In: Proceedings of the 2002 IEEE International Conference on Fuzzy Systems, pp. 1616–1622 (2002)Google Scholar
- 10.Barbosa, R., Belo, O.: Algorithmic Trading Using Intelligent Agents. In: Proceedings of the 2008 International Conference on Artificial Intelligence (2008)Google Scholar
- 11.Barbosa, R., Belo, O.: A Step-By-Step Implementation of a Hybrid USD/JPY Trading Agent. International Journal of Agent Technologies and Systems (2009)Google Scholar
- 12.Weka API, http://www.cs.waikato.ac.nz/ml/weka/
- 13.JBoss Drools API, http://www.jboss.org/drools/
- 14.ActiveMQ API, http://activemq.apache.org/