Abstract

We suggest a simple practical method to combine the human and artificialintelligence to both learn best investment practices of fund managers, andprovide recommendations to improve them. Our approach is based on a combinationof Inverse Reinforcement Learning (IRL) and RL. First, the IRL component learnsthe intent of fund managers as suggested by their trading history, and recoverstheir implied reward function. At the second step, this reward function is usedby a direct RL algorithm to optimize asset allocation decisions. We show thatour method is able to improve over the performance of individual fund managers.

 

Quick Read (beta)