Making Trading Decision with Machine Learning

This project works on optimizing liquidation strategy with reinforcement learning.

Currently, we are simulating the market environment with the Almgren-Chriss model and trying to explore the implementation of trading strategies.

Specifically, we are doing the following work:

1. Trying to implement the closed form solution under single agent settings
2. Using reinforcement techniques to train a single agent and compare its trading trajectory to the closed form one
3. Moving on to multi-agent settings and trying to explore the relationships between different agents

Content

• Market Simulation
• Optimal Liqudation Strategy
  • Closed Form Solution
  • Actor-Critic
  • Q-Network
• Model Extension
  • Multi-agent Environment

Market Simulation

Optimal Liqudation Strategy

Closed Form Solution

Actor-Critic

Q-Network

Model Extension

Multi-agent Environment