Sinergy Al Model Xgb Boost
- 专家
- Yevgeniy Koshtenko
- 版本: 1.0
- 激活: 5
This Al trading expert advisor (EA) uses two machine learning models - Chinese XGBoost and Russian CatBoost, working in pairs and enhancing each other's signals. The Al trading advisor is specifically designed for the euro-dollar pair and trained on data from the RoboForex broker. By using two models, the Al trading advisor can effectively analyze the market and provide more accurate trading signals.
The model is trained on data up to 2010, which allows it to effectively predict future prices. The forward test can be viewed after 2010, outside the training section. The Al trading model uses 99 features that enable it to effectively analyze the market and provide more accurate trading signals. According to testing since 2000, the expected annual return is around 30% per annum with a drawdown of 5% per annum. The average win rate is 83% profitable trades, the Sharpe ratio is over 7.
To choose the lot size, an improved Kelly formula is used, which takes the signal from one of the Al trading models. The stop, take profit, and risk sizes can be optimized according to personal preferences and risk level. The Al trading model overcomes the overfitting problem and in the future, new improved models will be presented for other assets.
For training the Al trading model, such technologies as cross-validation, feature engineering, feature clustering, data preprocessing were used. The Al trading model uses a hybrid of regression and classification models in one, which allows it to more effectively predict prices. Several new versions of the Al trading model are developed daily to achieve better results. In the future, it is planned to develop one model for each popular asset. The developer of the Al trading advisor is an author with over 8 years of experience in the market, an expert in financial markets.
For training the Al trading model, prices and their derivatives for the EURUSD pair were used, obtained from the RoboForex broker.