A production-grade machine learning trading system built on Binance Futures, featuring real-time signal generation, walk-forward backtesting, and an interactive Streamlit dashboard.
| Live Signal Dashboard | Backtest Equity Curve | Feature Importance |
|---|---|---|
 |
 |
 |
- ML Signal Engine — Random Forest classifier trained on 14 technical indicators, predicts next-bar direction with confidence scores
- Walk-Forward Cross-Validation — TimeSeriesSplit prevents data leakage, produces realistic out-of-sample accuracy
- Full Feature Pipeline — RSI, MACD, Bollinger Bands, ATR, Stochastic, OBV, EMA crossovers, volume z-scores
- Vectorised Backtesting — models transaction costs, slippage, drawdown; computes Sharpe, Sortino, win rate
- Interactive Dashboard — Streamlit app with live signal gauge, candlestick chart, equity curve, and order panel
- Risk Management Layer — pre-trade quantity limits, symbol allowlists, retry logic with exponential backoff
- Modular CLI —
signal,order, andbacktestsubcommands
alphapulse/
├── bot/
│ ├── client.py # Binance Futures API client with retry logic
│ ├── orders.py # OrderExecutor + RiskManager
│ ├── validators.py # Input sanitization
│ └── logging_config.py # Dual file+console logging
├── ml/
│ ├── features.py # OHLCV → 14-feature engineering pipeline
│ └── model.py # SignalModel (Random Forest / GBM)
├── backtest/
│ └── engine.py # Vectorised backtester + performance metrics
├── dashboard/
│ └── app.py # Streamlit 4-tab interactive dashboard
├── data/models/ # Persisted joblib model artifacts
├── logs/ # trading.log
├── cli.py # Unified CLI entrypoint
└── requirements.txt
Data flow:
Binance API → OHLCV klines → Feature Engineering → ML Model → Signal (BUY/SELL + confidence)
↓
OrderExecutor (with RiskManager)
↓
Binance Futures Testnet
Model: sklearn.ensemble.RandomForestClassifier
Input: 14 engineered features (see ml/features.py)
Target: Binary — 1 if next bar closes higher, 0 otherwise
Validation: 5-fold TimeSeriesSplit walk-forward CV
| Feature Group | Indicators |
|---|---|
| Returns | Log-return, % change |
| Momentum | RSI(14), MACD(12/26/9), Stochastic %K/%D |
| Trend | EMA crossover(9/21), Bollinger Band position |
| Volatility | ATR%, Bollinger Band width |
| Volume | OBV change, Volume z-score |
| Metric | Definition |
|---|---|
| Sharpe Ratio | Annualised risk-adjusted return |
| Sortino Ratio | Downside deviation adjusted return |
| Max Drawdown | Largest peak-to-trough loss |
| Win Rate | % of trades closed profitably |
| Profit Factor | Gross profit / Gross loss |
| ROC-AUC | ML classifier discrimination ability |
# 1. Clone
git clone https://github.com/yourusername/alphapulse.git
cd alphapulse
# 2. Install dependencies
pip install -r requirements.txt
# 3. Configure environment
cp .env.example .env
# Edit .env and add your Binance Testnet API keys.env.example:
BINANCE_API_KEY=your_testnet_api_key
BINANCE_API_SECRET=your_testnet_api_secret
streamlit run dashboard/app.pypython cli.py signal --symbol BTCUSDT --interval 1h --lookback 500python cli.py backtest --symbol BTCUSDT --interval 1h --capital 10000# Market order
python cli.py order --symbol BTCUSDT --side BUY --type MARKET --quantity 0.01
# Limit order
python cli.py order --symbol ETHUSDT --side SELL --type LIMIT --quantity 0.1 --price 3200pytest tests/ -v --cov=.- LLM Sentiment Analysis — Integrate GPT-4 / FinBERT on live crypto news from NewsAPI to generate a sentiment score as an additional feature
- Reinforcement Learning Agent — Replace the RF classifier with a PPO/DQN agent (Stable Baselines3) for end-to-end policy learning
- Portfolio Optimisation — Multi-asset mean-variance optimisation with dynamic position sizing using PyPortfolioOpt
- Live Deployment — Docker + cron job for scheduled signal refresh; webhook alerts via Telegram Bot API
- Model Drift Detection — Monitor feature distributions with Evidently AI; retrain automatically when drift is detected
| Layer | Tools |
|---|---|
| Language | Python 3.11 |
| Exchange API | python-binance |
| ML | scikit-learn (Random Forest, GBM) |
| Feature Engineering | pandas, numpy |
| Backtesting | Custom vectorised engine |
| Dashboard | Streamlit, Plotly |
| Model Persistence | joblib |
| Testing | pytest, pytest-cov |
| Logging | Python logging (file + console) |
Built by [So-coder-ai] | GitHub | LinkedIn
MIT License — free to use and modify.
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
# Clone and setup
git clone https://github.com/So-coder-ai/alphapulse.git
cd alphapulse
pip install -r requirements.txt
# Configure API keys
cp .env.example .env
# Edit .env with your Binance Testnet API keys
# Run the dashboard
streamlit run dashboard/app.py