A Graphical Model for Collective Behavior Learning Using Minority Games
The Minority Game (MG) is a simple game theory model for the collective behavior of agents in an idealized situation where they compete for some finite resource. In this paper, we assume that collective behavior is determined by the aggregation of individual actions of agents. This causal relation between collective behavior and individual actions is investigated. A graphical model is proposed to model the generative process of collective behavior using a group of agents whose actions are modeled by minority games. In this model, we can infer the individual behavior of the agents by training on the global information, and then make predictions about the future collective behavior. Experimental results on a set of stock indexes from the Chinese market and foreign exchange (FX) rates show that the new proposed model can effectively capture the rises and falls of market and be significantly better than a random predictor. This framework also provides a new data mining paradigm for analyzing collective data by modeling micro-level actions of agents using game theory models.
KeywordsCollective Intelligence Minority Game Probabilistic Graphical Models
Unable to display preview. Download preview PDF.
- 1.Rapoport, A., Chammah, A., Orwant, C.: Prisoner’s Dilemma: A Study in Conict and Cooperation. University of Michigan Press, Ann Arbor, Michigan (1965)Google Scholar
- 5.Johnson, N.F., Jeffries, P., Hui, P.M.: Financial market complexity. Oxford University Press, Oxford (2003)Google Scholar
- 6.Qin, Z.: Market mechanism designs with heterogeneous trading agents. In: Proceedings of Fifth International Conference on Machine Learning and Applications (ICMLA 2006), pp. 69–74 (2006)Google Scholar
- 9.Arthur, W.B.: Bounded rationality and inductive behavior (the el farol problem). American Economic Review 84, 406–411 (1994)Google Scholar
- 11.Challet, D., Marsili, M., Zhang, Y.C.: Minority Games: Interacting Agents in Financial Markets. Oxford University Press, USA (2004)Google Scholar
- 17.Koller, D., Friedman, N.: Probabilistic Graphical Models: Principles and Techniques. MIT Press (2009)Google Scholar
- 19.Minka, T., Winn, J., Guiver, J., Knowles, D.: Infer.NET 2.4, Microsoft Research Cambridge (2010), http://research.microsoft.com/infernet