💲 AI/ML Trading Strategies with XGBoost

An institutional-grade, educational application that applies XGBoost machine learning to historical market data for multi-class trading signal generation. The system computes 33+ technical indicators, auto-selects the best labelling strategy, evaluates model performance, explains predictions with SHAP, and backtests the resulting signals.

Link to live terminal - https://tradewithmachine.streamlit.app/

⚠️ Disclaimer: This project is for educational and informational purposes only. It is not financial advice. Always do your own research before making investment decisions.

---

Features

Core ML Pipeline

• Data Acquisition – Pulls OHLCV data from Yahoo Finance for Indian indices, futures, and crypto assets.
• Feature Engineering – Builds 33+ technical indicators across momentum, oscillators, trend, volatility, and volume categories.
• Automated Labelling – Generates bullish / bearish / neutral labels across multiple lookahead windows and thresholds.
• Baseline Scan – Trains a quick XGBoost model for every label strategy and ranks them by macro-F1.
• Hyper-Parameter Optimization – Optional GridSearchCV with time-series cross-validation.
• Walk-Forward Validation – Optional out-of-sample validation across multiple time folds.

Model Interpretability & Evaluation

• Classification Reports – Accuracy, balanced accuracy, macro-F1, precision, recall, MCC.
• Confusion Matrix – Normalised heatmap of predicted vs actual classes.
• ROC Curves – One-vs-rest ROC/AUC for each class.
• SHAP Summary – Global feature importance to understand what drives predictions.

Signals & Backtesting

• Probability Thresholding – User-defined bullish / bearish thresholds to reduce false signals.
• Live Trading Signal – Latest prediction with confidence breakdown.
• Signal Timeline Chart – Candlestick chart overlaid with BUY / SELL markers.
• Strategy Backtest – Compares cumulative strategy return vs buy-and-hold market return.
• Risk Metrics – Win rate, alpha, total signals, and market return.

UI/UX

• Modern dark-themed interface using custom design tokens (theme.py).
• Reusable chart components (components.py) and styled cards (ui_styles.py).
• Landing page with feature overview and clear call-to-action.

---

Tech Stack

Layer	Libraries
App Framework	Streamlit
Data	yfinance, pandas, numpy
Technical Indicators	pandas-ta
Machine Learning	XGBoost, scikit-learn
Explainability	SHAP
Visualisation	Plotly, Matplotlib, Altair

---

Installation

1. Clone or navigate to the project folder:

   cd "XGboost ML"

2. Create a virtual environment (recommended):

   python -m venv venv
   # Windows
   venv\Scripts\activate
   # macOS / Linux
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

---

Usage

Run the Streamlit application from the project root:

streamlit run p1.py

Then open your browser at http://localhost:8501.

Workflow

1. Select an asset symbol in the sidebar (e.g. ^NSEI, ^BSESN, BTC-USD).
2. Set the start date and lookback window.
3. Choose a labelling strategy or leave it on auto.
4. Adjust bullish / bearish probability thresholds.
5. Optionally enable Hyper-Parameter Search or Walk-Forward Validation.
6. Click Execute Strategy to train the model and view results.

---

Model Details

• Algorithm: XGBClassifier with multi:softprob objective for 3-class classification.
• Classes:
  • 0 – Neutral
  • 1 – Bearish
  • 2 – Bullish
• Default Hyperparameters:
  • n_estimators=600
  • max_depth=4
  • learning_rate=0.01
  • subsample=1.0
  • colsample_bytree=0.6
  • tree_method='hist'
• Data Split: 60% train / 20% validation / 20% test (chronological).

---

📊 Indicators Used

Category	Indicators
Momentum	RSI(5, 10, 15), ROC(10), MOM(10)
Oscillators	Stochastic RSI, CCI(20), Williams %R(14), KST, MACD
Trend	SMA(5, 10, 20), EMA(5, 10, 20), VWMA(20)
Volatility	Bollinger Bands, ATR(14), Keltner Channel
Volume	OBV, A/D Line, EFI, NVI, PVI

---

Author

Engineer
📧 contact: ggengineerco@gmail.com

---

📜 License

This project is released for educational purposes. Use at your own risk.