Stock trading rule discovery with double deep Q-network

Stock market serves as an important indicator of today’s economy. Predicting the price fluctuation stocks and acquiring maximum gains has been main concern investors. In recent years, deep learning models are widely applied to stock prediction have achieved good performances. However, majority these based belong supervised methods not capable dealing with long-term targets. Therefore, in this paper we proposed a reinforcement trading model, which is suitable for predicting transactions. We carefully devise reward function policy network, enables model capture hidden dependencies latent dynamics data. order evaluate superiority trend forecasting transaction conducted on randomly selected indices. Experiment results demonstrate that our outperforms baseline several indicators.