A novel method of modeling dynamic evolutionary game with rational agents for market forecasting


Talebimotlagh N., Hashemzadeh F., Rikhtehgar Ghiasi A., Ghaemi S.

Economic Computation and Economic Cybernetics Studies and Research, vol.51, no.1, pp.281-302, 2017 (SCI-Expanded) identifier

  • Publication Type: Article / Article
  • Volume: 51 Issue: 1
  • Publication Date: 2017
  • Journal Name: Economic Computation and Economic Cybernetics Studies and Research
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI), Scopus
  • Page Numbers: pp.281-302
  • Keywords: Evolutionary game theory, Evolutionary stable state, Gold market, Rational agent, Recurrent neural network, Reinforcement learning, Two step ahead prediction
  • Istanbul Technical University Affiliated: No

Abstract

Gold price modeling and prediction is a difficult problem and drastic changes of the price causes nonlinear dynamic that makes the price prediction one of the most challenging tasks for economists. Since gold market always has been interesting for traders, many of traders with various beliefs were highly active in gold market. The competition among two agents of traders, namely trend followers and rational agents, to gain the highest profit in gold market is formulated as a dynamic evolutionary game, where, the evolutionary equilibrium is considered to be the solution to this game. Furthermore, genetic algorithm is being used to find the unknown parameters of the model, so that we could maximize the fitness of the proposed multi agent model and the gold market daily price data. Besides the evolutionary game dynamic, we proposed a new method for modeling rational expectations using recurrent neural network. The evolutionarily stable strategies is proven despite the prediction error of the expectation. The empirical results show the high efficiency of the proposed method which could forecast future gold price precisely.