Portfolio Optimization Under the Framework of Reinforcement Learning

Abstract

Portfolio management is an art and science of making decisions about the investment mix and policy, that matching investments to objectives, assets allocation for the individuals and institutions, balancing risk against performance. In this paper, a neural network is used to train and analysis the history of the assets as well as their portfolio weights for each asset and later evaluate the potential growth for the immediate future under the reinforcement learning(RL) framework, which is based on the recent reinforcement learning(RL) developments, testing is carried with the historical data from cryptocurrency exchange market, camparison with other framework shows that the framework based on RL behaved far better than most other optimization framework in the test period.