SimTradeML is a modular, cloud-ready framework for training, evaluating, and deploying machine learning models in financial simulation environments. It serves as the predictive engine behind the SimTrade ecosystem, bridging raw data from SimTradeData and strategy logic in SimTradeLab.

SimTradeML provides reusable ML pipelines and model services tailored for financial workflows. It enables:

• Time series forecasting, classification, and regression for market signals
• Model packaging as RESTful APIs for strategy integration
• Cloud-native deployment via AWS CDK and GitLab CI
• Feedback loops between strategy performance and model retraining

simtradeML/
├── forecast_engine/      # ML models: ARIMA, Prophet, LSTM, XGBoost, etc.
├── model_api/            # FastAPI endpoints for model inference
├── data_ingest/          # ETL pipelines consuming SimTradeData outputs
├── infra/                # AWS CDK / Terraform templates for deployment
├── ci_pipeline/          # GitLab CI/CD scripts
├── notebooks/            # Exploratory analysis and prototyping
├── tests/                # Unit and integration tests
└── README.md

SimTradeML is designed to work seamlessly with other SimTrade modules:

git clone https://github.com/ykayz/SimTradeML.git
cd SimTradeML

pip install -r requirements.txt

python forecast_engine/train_xgboost.py

uvicorn model_api.main:app --reload

Train a volatility prediction model using ETF data from SimTradeData, deploy it as an API, and call it from a SimTradeLab strategy to generate dynamic position sizing signals.

• Python, scikit-learn, XGBoost, statsmodels
• FastAPI, Docker, AWS CDK
• GitLab CI/CD
• Streamlit (optional dashboard)
• Jupyter Notebooks for prototyping

Actively evolving. Initial modules include:

• ETF volatility prediction
• Trend classification
• Strategy-model feedback loop prototype

Contributions are welcome! Please submit issues or pull requests for:

• New model types
• Deployment templates
• Strategy integration examples
• Documentation improvements

MIT License. See LICENSE for details.