Optimal Order Execution Based On Reinforce Learning

The problem I am trying to solve in this paper is optimal order execution problem in exchange algorithm trading.At first,we reviewed the three generations order execution algorithm.The first-generation is uncertainty bond model,the second-generation model is implement shortfall model,the third-generation is stochastic control model.By relaxing the parameter assumption about market process,we modeled the optimal order execution problem into a stochastic control problem with constraint.By using reinforce learning with high frequency data,we get an optimal order execution strategy which is consist with real market.The result of the paper tells us liquidity,order execution,aggressive order amount has significant market information.The optimal order execution strategy has less trading cost than one shot optimal order execution.