Abstract

According to documents, there has not been a completely artificialintelligence framework with shorting mechanism in continuous action space forportfolio optimization. The objective of this paper is to verify that currentcutting-edge technology, deep reinforcement learning, can be applied inportfolio management and help us get artificial intelligence. We improve on theexisting Deep Reinforcement Learning Portfolio model and make many innovations,such as adding short mechanism, designing an arbitrage mechanism. Inexperiments, we use our model in several randomly selected portfolios, theresults show we can get artificial intelligence and defeat market through deepreinforcement learning.

 

Quick Read (beta)