CryptoPredict 🔮

AI-powered cryptocurrency price forecasting with an active learning feedback loop.

---

✨ Features

Feature	Description
🧠 LSTM Model	Multi-layer LSTM trained on historical price data with RSI, Bollinger Bands, and MA indicators
📈 7/14-day Forecast	Autoregressive multi-step price prediction with confidence scoring
💬 Active Learning	User feedback loop that tracks model accuracy and triggers retraining
📊 Live Metrics	MAE, RMSE, and accuracy dashboard updated in real time
🔄 Auto-Retrain	Background retraining triggered after every 50 feedback entries
🌐 CoinGecko API	Free-tier market data — no API key required
🐳 Docker	Single-command full-stack deployment

---

📸 Screenshots

Add screenshots here after running the app locally.

---

🏗 Architecture

cryptopredict/
├── backend/
│   ├── main.py              # FastAPI application & REST endpoints
│   ├── train_model.py       # LSTM training pipeline (run once before API)
│   ├── requirements.txt
│   ├── Dockerfile
│   └── tests/
│       └── test_api.py
├── frontend/
│   ├── src/
│   │   ├── App.jsx          # Main React application
│   │   ├── utils/api.js     # Typed API client
│   │   ├── hooks/           # Custom React hooks
│   │   └── index.css        # Global styles
│   ├── Dockerfile
│   ├── nginx.conf
│   └── vite.config.js
├── docker-compose.yml
└── .github/workflows/ci.yml

---

🚀 Quick Start

Prerequisites

• Python 3.11+
• Node.js 20+
• (Optional) Docker & Docker Compose

---

Option 1 — Docker (recommended)

git clone https://github.com/Sm3th/cryptopredict.git
cd cryptopredict

# 1. Train the model first (runs outside Docker to save the .keras file)
cd backend
pip install -r requirements.txt
python train_model.py --coin bitcoin --days 730

# 2. Spin up everything
cd ..
docker-compose up --build

Open http://localhost for the app, http://localhost:8000/docs for Swagger UI.

---

Option 2 — Local development

Backend

cd backend
python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate
pip install -r requirements.txt

# Train the model (required before starting the API)
python train_model.py --coin bitcoin --days 730

# Start API
uvicorn main:app --reload --port 8000

Frontend

cd frontend
npm install
npm run dev
# → http://localhost:5173

---

🧠 Model Details

Architecture

Input (60 × 1)
   ↓
LSTM(128) → Dropout(0.2)
   ↓
LSTM(64)  → Dropout(0.2)
   ↓
LSTM(32)  → Dropout(0.2)
   ↓
Dense(16, relu)
   ↓
Dense(1)  ← price prediction

Features used

Feature	Description
price	Daily closing price (normalized 0–1)
MA_7 / MA_21 / MA_50	Moving averages
RSI	Relative Strength Index (14-period)
BB_upper / BB_lower	Bollinger Bands (20-period, ±2σ)
volatility	7-day rolling std
momentum	price − price[−10]

Training configuration

Parameter	Value
Lookback window	60 days
Train / Test split	80 / 20
Epochs	50 (early stopping, patience=10)
Batch size	32
Optimiser	Adam
Loss	MSE

---

🔌 API Reference

Method	Endpoint	Description
GET	/	Health check
GET	/api/health	Detailed system status
POST	/api/predict	Run price forecast
POST	/api/feedback	Submit prediction feedback
POST	/api/retrain	Trigger background retraining
GET	/api/metrics	Model performance metrics
GET	/api/feedback/history	All feedback records
GET	/api/coins	Supported cryptocurrencies

Full interactive documentation: http://localhost:8000/docs

Example — Predict

curl -X POST http://localhost:8000/api/predict \
  -H "Content-Type: application/json" \
  -d '{"coin_id": "bitcoin", "days": 7}'

{
  "coin_id": "bitcoin",
  "current_price": 67450.00,
  "predictions": [
    { "day": 1, "date": "2026-05-28", "price": 68120.50, "change_percentage": 0.99 },
    ...
  ],
  "confidence": 0.70,
  "sentiment": "Bullish",
  "sentiment_score": 72.4,
  "model_accuracy": 68.0
}

Example — Feedback

curl -X POST http://localhost:8000/api/feedback \
  -H "Content-Type: application/json" \
  -d '{
    "prediction_id": "bitcoin_1234567890",
    "actual_price": 68000.00,
    "predicted_price": 68120.50,
    "date": "2026-05-28",
    "is_accurate": true
  }'

---

🧪 Tests

cd backend
pip install pytest httpx anyio
pytest tests/ -v

---

⚙️ Environment Variables

Variable	Default	Description
ALLOWED_ORIGINS	http://localhost:5173,...	CORS allowed origins
MODEL_PATH	crypto_lstm_model.keras	Path to trained model
SCALER_PATH	scaler.pkl	Path to fitted scaler
VITE_API_URL	/api	API base URL (frontend)

---

🗺 Roadmap

• Multi-coin comparison view
• Sentiment analysis from Twitter/Reddit
• Email alerts for large predicted swings
• Model versioning & rollback
• PostgreSQL for persistent feedback storage

---

⚠️ Disclaimer

This project is built for educational purposes only. Cryptocurrency markets are highly volatile. Do not make financial decisions based on these predictions.

---

📄 License

MIT © 2026 — İsmet Organ