Developing an Investment Method for Securities With Reinforcement Learning

In order to overcome the uncertainty of the securities market environment, in this paper we propose an investment method based on Reinforcement Learning (RL) with combining different behavioral preferences. The investment method is achieved by using Support Vector Machine(SVM) to classify the agents with two preference categories. We select the trading mode that deviates the farthest from the classification boundary in each step and enhances the objectivity of comprehensive trading strategies, especially in the face of environmental uncertainty. The effectiveness of the investment method is validated by comparing with various RL methods from multiple securities markets.