Ample research mainly based on supervised learning-oriented approaches already has been done in the real time financial series forecasting area. Although some of such predictors can perform better than conventional models consistently, they suffer from inherent drawbacks due to the characteristic of supervised learning. Reinforcement Learning (RL) methods embody a general Monte Carlo approach to dynamic programming for catching system dynamics under uncertain environments. The integrated architecture combined RL and artificial neural networks (ANNs) is the powerful tool for both control and prediction purpose. Relatively fewer studies were made in terms of predicting the stock market data by means of such hybrid framework. This paper presents the novel Neuro - Q Learning prediction model for forecasting real time stock index. The encouraging experimental results for NASDAQ market illustrate the effectiveness of the presented approach.
展开▼