This project is for fun! And profit, if any.
The goal of this project is :
- Develope an alpha model predicting Forex Movements (with accurary >= 70%);
- Integrate it with a Algo Trading Enginne;
- Add on Risk Management Modules to control drawdowns (e.g. Position Sizer & Emergency Stopper);
EURUSD 5-minute price data is from Dukascopy.com using their historical data feed.
- Lots of price-based features added;
- Used GDBT for benchmark prediction (no feature selection & parameters adjustments), got accuracy score of 0.6690;
- A simple backtest example using Backtrader(an Open Source framework for backtest & live trading);
- Deep Neural Network model prediction (classification);
- Change from a classification to a regression model to produce a profit target or a signal strength indicator;
- Integrate model with Backtrader framework;
- Add leverage/margin/commission setting;
- Add position sizer;