CryptoPredict

AI-powered cryptocurrency analysis platform that combines LSTM neural networks, FinBERT sentiment analysis, and real-time market data to predict crypto price movements and let users test their market intuition.

---

Features

• Real-Time Dashboard — Live OHLC charts with MACD indicators and sparklines for BTC, ETH, BNB, XRP & SOL
• AI Price Prediction — LSTM model trained on news sentiment and technical indicators to predict 60-minute price returns
• FinBERT Sentiment Analysis — Automatic sentiment scoring of crypto news headlines using a fine-tuned BERT model
• Prediction Game — Interactive click-to-predict game where users guess future prices and earn accuracy-based points
• Leaderboard — Compete with other players and track prediction accuracy rankings
• Live News Feed — Real-time cryptocurrency news with auto-rotation

---

Tech Stack

Layer	Technology
Frontend	React 19, Material UI 7, Recharts, Axios
Backend	Python, Flask, SQLAlchemy, Flask-CORS
ML/AI	PyTorch LSTM, FinBERT (HuggingFace), Scikit-learn GBR
APIs	Binance API (market data), NewsAPI (news)
Database	SQLite

---

Project Structure

Crypto_Prediction_Application/
├── app/                          # Flask backend
│   ├── __init__.py               # App factory with dotenv loading
│   ├── models.py                 # SQLAlchemy models (User, Prediction)
│   ├── routes.py                 # API endpoints
│   └── utils.py                  # LSTM model loading, Binance/NewsAPI clients
├── frontend/                     # React frontend
│   └── src/
│       ├── components/           # UI components
│       │   ├── Charts.js         # Main dashboard with charts & market data
│       │   ├── Home.js           # Landing page
│       │   ├── Login.js          # Authentication
│       │   ├── Register.js       # User registration
│       │   ├── QuizPage.js       # Interactive prediction game
│       │   ├── Leaderboard.js    # Points ranking
│       │   └── Navbar.js         # Navigation bar
│       ├── context/              # React context (auth state)
│       ├── utils/                # Client-side utilities
│       └── theme.js              # Shared MUI dark theme
├── scripts/                      # Data pipeline scripts
│   ├── build_dataset.py          # Fetch news + OHLC data from APIs
│   ├── csv_preprocess.py         # Feature engineering
│   ├── data_split.py             # Train GBR model
│   └── sentiment_scores.py       # FinBERT sentiment analysis
├── LSTM_model.py                 # LSTM model definition + training
├── run.py                        # Backend entry point
├── requirements.txt              # Python dependencies
└── .env.example                  # API key template

---

Prerequisites

• Python 3.8+
• Node.js 16+ and npm
• API keys for Binance and NewsAPI

---

Installation

1. Clone the Repository

git clone https://github.com/muradhuseynov1/Crypto_Prediction_Application.git
cd Crypto_Prediction_Application

2. Backend Setup

# Create virtual environment
python -m venv .venv

# Activate it
# Windows:
.venv\Scripts\activate
# macOS/Linux:
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Set up environment variables
cp .env.example .env
# Edit .env with your actual API keys

3. Frontend Setup

cd frontend
npm install
cd ..

---

Running the Application

Open two terminals:

Terminal 1 — Backend:

python run.py

Backend runs at http://localhost:5000

Terminal 2 — Frontend:

cd frontend
npm start

Frontend runs at http://localhost:3000

---

API Keys

Create a .env file in the project root (use .env.example as template):

BINANCE_API_KEY=your_binance_api_key
BINANCE_API_SECRET=your_binance_api_secret
NEWSAPI_KEY=your_newsapi_key
COINLAYER_API_KEY=your_coinlayer_api_key

See API_KEYS_GUIDE.md for detailed instructions on obtaining keys and storing them securely.

---

Data Pipeline

The ML models were trained on a custom dataset built from crypto news and price data. To rebuild from scratch:

# 1. Collect news + OHLC data
python scripts/build_dataset.py

# 2. Add FinBERT sentiment scores
python scripts/sentiment_scores.py

# 3. Engineer technical features
python scripts/csv_preprocess.py

# 4. Train GBR model
python scripts/data_split.py

# 5. Train LSTM model
python LSTM_model.py

---

API Endpoints

Method	Endpoint	Description
POST	/api/register	Register a new user
POST	/api/login	Authenticate user
POST	/api/predict	Submit a price prediction
GET	/api/leaderboard	Get top 10 players
GET	/api/charts/<symbol>	Historical OHLC + MACD data
GET	/api/market-overview	24h price changes for all coins
GET	/api/news	Latest crypto news

---

License

MIT License — see LICENSE.

This project was built as part of the Revolut challenge during HackUPC 2025 hackathon by Murad Hüseynov and Giorgia Barboni.