This repository contains code to train a Deep Reinforcement Learning (DRL) Agent to trade stocks. The trained model is connected to a stock broker platform to make real trades in real time, using real money. FinRL was used at the core for the implementation.
The Agent only trades a single time a day (at the end), with the goal of maximizing profits. The Agent is encouraged to buy assets instead of holding cash by introducing a cash penalty.
- Fetch current stock information (real time)
- Connect with De Giro API
- Link model with De Giro to make trades
- Set everything on a cloud server