Multi-agent Investment Management With Emotional Learning

The investment domain, especially stock market is a dynamically changing, stochastic and unpredictable environment. In this market there is a lot of information about the stocks and market conditions. If a capital investment system uses this information, it can select a portfolio more efficiently. Modern theories of portfolio selection [12] try to provide the best possible expected rate of return for a specified level of risk.However, they make some presumption, and if they don't hold, these methods are no longer efficient. More recently, AI techniques have been used in several papers proposing less model- based approaches. Reinforcement learning algorithm is among model- free techniques successfully used in prediction and decision making [14, 20]. The main contribution of this paper is to provide a multi-agent system employing reinforcement learning with emotional signal for decisionmaking in portfolio selection that gathers information, news and specialist analysis to make decision for optimal investment in stock market.

Keywords:

Multi-Agent system , Q-Learning , Investment , Emotional Learning