🤖 AI Trading Assistant

A comprehensive, AI-powered trading assistant for stock and cryptocurrency markets that combines technical analysis, sentiment analysis, and machine learning to provide intelligent trading signals with detailed explanations.

🎯 Project Overview

This AI Trading Assistant is designed to be both a powerful trading tool and an educational platform. When it generates BUY/SELL/HOLD signals, it provides comprehensive explanations of the reasoning behind each decision, helping users understand:

• Technical Analysis: Moving averages, RSI, MACD, Bollinger Bands, and more
• Market Sentiment: News sentiment analysis using FinBERT
• AI Predictions: LSTM neural network price forecasting
• Risk Assessment: Confidence-based risk level classification

🏗️ Architecture

The application consists of three core modules:

1. Data Pipeline (data_collector.py)

• Market Data: Fetches OHLCV data from yfinance (stocks) and Binance API (cryptocurrencies)
• News Data: Web scraping from financial news sources (Yahoo Finance, Google News)
• Real-time Updates: Configurable update frequency

2. AI & Machine Learning (prediction_model.py, sentiment_analysis.py)

• Sentiment Analysis: FinBERT model for financial text sentiment classification
• Price Prediction: LSTM neural networks for time series forecasting
• Feature Engineering: 25+ technical indicators and advanced feature selection
• Model Training: Automated training with validation and performance metrics

3. Educational Engine (trading_bot.py)

• Signal Generation: Combines technical and sentiment analysis
• Detailed Explanations: Human-readable reasoning for every signal
• Risk Assessment: Confidence-based risk classification
• Portfolio Management: Multi-symbol tracking and analysis

🚀 Features

Core Capabilities

• Multi-Asset Support: Stocks and cryptocurrencies
• Real-time Analysis: Live market data and news sentiment
• Technical Indicators: 25+ advanced technical indicators
• AI Predictions: LSTM-based price forecasting
• Sentiment Analysis: FinBERT-powered news sentiment
• Educational Insights: Detailed explanations for every signal

User Interface

• Web Dashboard: Flask-based web application
• Streamlit Alternative: Modern, interactive Streamlit app
• Responsive Design: Mobile-friendly interface
• Real-time Updates: Live signal refresh capabilities

Data Management

• Export/Import: Save and load trading signals
• Historical Data: Configurable time periods
• Portfolio Tracking: Multi-symbol management
• Performance Metrics: Training and testing statistics

📋 Requirements

System Requirements

• Python 3.9+
• 8GB+ RAM (for LSTM training)
• Stable internet connection
• GPU recommended (for faster model training)

Python Dependencies

pandas==2.1.4
numpy==1.24.3
yfinance==0.2.28
requests==2.31.0
beautifulsoup4==4.12.2
scikit-learn==1.3.2
tensorflow==2.15.0
transformers==4.36.2
torch==2.1.2
flask==3.0.0
plotly==5.17.0
streamlit==1.29.0
python-binance==1.0.19
ta==0.10.2
nltk==3.8.1
textblob==0.17.1

🛠️ Installation

1. Clone the Repository

git clone <repository-url>
cd AI_Trading

2. Create Virtual Environment

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Download NLTK Data

import nltk
nltk.download('punkt')
nltk.download('stopwords')
nltk.download('vader_lexicon')

🚀 Usage

Quick Start

1. Basic Usage (Python Script)

from trading_bot import AITradingBot

# Initialize the bot
bot = AITradingBot()

# Add symbols to track
bot.collect_market_data('AAPL', period='6mo')
bot.collect_news_data('AAPL', limit=20)

# Generate trading signal
signal = bot.generate_trading_signal('AAPL')
print(f"Signal: {signal['signal']}")
print(f"Confidence: {signal['confidence']:.1%}")
print(f"Explanation: {signal['explanation']}")

2. Web Dashboard (Flask)

python app.py

Open your browser and navigate to http://localhost:5000

3. Streamlit App

streamlit run streamlit_app.py

Advanced Usage

Portfolio Analysis

# Analyze multiple symbols
symbols = ['AAPL', 'GOOGL', 'MSFT', 'TSLA']
portfolio_signals = bot.get_portfolio_signals(symbols)

for signal in portfolio_signals:
    print(f"{signal['symbol']}: {signal['signal']} ({signal['confidence']:.1%})")

Custom Configuration

config = {
    'lookback_days': 60,
    'technical_confidence_weight': 0.7,
    'sentiment_confidence_weight': 0.3,
    'n_selected_features': 30
}

bot = AITradingBot(config=config)

Model Training and Evaluation

# Train LSTM model
features, targets = bot.lstm_predictor.prepare_data(market_data)
history = bot.lstm_predictor.train(features, targets)

# Evaluate performance
test_metrics = bot.lstm_predictor.evaluate(test_features, test_targets)
print(f"R² Score: {test_metrics['r2']:.4f}")

📊 Technical Indicators

The system calculates over 25 technical indicators:

Trend Indicators

• Simple Moving Averages (SMA 5, 10, 20, 50, 200)
• Exponential Moving Averages (EMA 12, 26, 50)
• MACD (Moving Average Convergence Divergence)
• Parabolic SAR
• ADX (Average Directional Index)
• Golden Cross/Death Cross detection

Momentum Indicators

• RSI (Relative Strength Index)
• Stochastic Oscillator
• Williams %R
• ROC (Rate of Change)
• CCI (Commodity Channel Index)
• Money Flow Index

Volatility Indicators

• Bollinger Bands (Upper, Middle, Lower, Width, Position)
• ATR (Average True Range)
• Keltner Channels

Volume Indicators

• Volume SMA
• On Balance Volume (OBV)
• Volume Rate of Change
• Chaikin Money Flow
• VWAP (Volume Weighted Average Price)

Support/Resistance

• Pivot Points
• Support/Resistance levels (R1, S1, R2, S2)
• Price position relative to levels

🧠 AI Models

LSTM Neural Network

• Architecture: Multi-layer LSTM with dropout and batch normalization
• Input: Time series sequences of technical indicators
• Output: Price predictions for next 1-7 days
• Training: Adam optimizer with early stopping and learning rate reduction
• Validation: 20% test split with performance metrics

FinBERT Sentiment Analysis

• Model: Pre-trained FinBERT for financial text
• Fallback: TextBlob for cases where FinBERT fails
• Features: Financial keyword extraction and market term identification
• Output: Positive/Negative/Neutral sentiment with confidence scores

Feature Selection

• Methods: Correlation, PCA, and variance-based selection
• Configurable: Number of features (default: 25)
• Dynamic: Adapts to available data and indicators

📈 Signal Generation

Signal Types

• BUY: Strong bullish indicators with positive sentiment
• SELL: Strong bearish indicators with negative sentiment
• HOLD: Mixed signals or insufficient confidence

Confidence Levels

• High (80%+): Strong technical and sentiment alignment
• Medium (60-79%): Good alignment with some uncertainty
• Low (<60%): Mixed or weak signals

Risk Assessment

• Low Risk: High confidence signals (>80%)
• Medium Risk: Medium confidence signals (60-80%)
• High Risk: Low confidence signals (<60%)

🔧 Configuration

Bot Configuration

default_config = {
    'lookback_days': 30,                    # Days to look back for analysis
    'prediction_days': 1,                   # Days to predict ahead
    'technical_confidence_weight': 0.6,     # Weight for technical analysis
    'sentiment_confidence_weight': 0.4,     # Weight for sentiment analysis
    'min_confidence_threshold': 0.6,        # Minimum confidence for signals
    'update_frequency_hours': 4,            # Data update frequency
    'max_news_articles': 20,                # Maximum news articles to analyze
    'feature_selection_method': 'correlation', # Feature selection method
    'n_selected_features': 25               # Number of features to select
}

Model Hyperparameters

lstm_config = {
    'lstm_units': [128, 64, 32],           # LSTM layer sizes
    'dropout_rate': 0.2,                    # Dropout for regularization
    'learning_rate': 0.001,                 # Adam optimizer learning rate
    'batch_size': 32,                       # Training batch size
    'epochs': 100,                          # Maximum training epochs
    'validation_split': 0.2                 # Validation data split
}

📱 Web Interface

Flask Dashboard

• Home: Project overview and quick start guide
• Dashboard: Symbol management and signal display
• Analysis: Detailed technical and sentiment analysis
• Settings: Bot configuration and data management
• About: Project information and disclaimers

Streamlit Alternative

• Modern UI: Clean, responsive interface
• Interactive Charts: Plotly-powered visualizations
• Real-time Updates: Live data refresh capabilities
• Mobile Friendly: Responsive design for all devices

📊 Data Sources

Market Data

• Stocks: Yahoo Finance (yfinance library)
• Cryptocurrencies: Binance API
• Intervals: 1m, 5m, 15m, 30m, 1h, 1d, 1w, 1m
• Periods: 1d, 5d, 1mo, 3mo, 6mo, 1y, 2y, 5y, 10y, ytd, max

News Sources

• Yahoo Finance: Financial news and market updates
• Google News: General financial news
• Real-time: Continuous news monitoring
• Sentiment: AI-powered sentiment analysis

🚨 Disclaimer

⚠️ IMPORTANT: This tool is for educational purposes only.

• Not Financial Advice: This tool does not provide financial advice
• Do Your Own Research: Always conduct thorough research before trading
• Consult Professionals: Consider consulting with a financial advisor
• Risk Warning: Trading involves risk of loss
• Past Performance: Past performance doesn't guarantee future results
• Educational Tool: Use as a learning platform for trading education

🤝 Contributing

We welcome contributions! Please see our contributing guidelines:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

Areas for Contribution

• Additional technical indicators
• New sentiment analysis models
• Enhanced visualization features
• Performance optimizations
• Documentation improvements

📚 Learning Resources

This tool is designed to help you learn about:

• Technical Analysis: Understanding market indicators and patterns
• Market Sentiment: How news affects market movements
• Machine Learning: AI applications in finance
• Risk Management: Portfolio analysis and risk assessment
• Trading Education: Practical application of financial concepts

🔍 Troubleshooting

Common Issues

1. Model Training Fails

# Check available memory
free -h  # Linux/Mac
# or
wmic computersystem get TotalPhysicalMemory  # Windows

# Reduce batch size or lookback days
config = {'batch_size': 16, 'lookback_days': 20}

2. News Scraping Issues

# Use alternative news sources
news_collector = NewsCollector()
news = news_collector.scrape_google_news("AAPL")

3. Memory Issues

# Reduce feature count
config = {'n_selected_features': 15}

4. API Rate Limits

# Add delays between requests
import time
time.sleep(1)  # 1 second delay

Performance Optimization

• Use GPU for LSTM training
• Reduce lookback period for faster analysis
• Limit number of tracked symbols
• Use feature selection to reduce dimensionality

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• FinBERT: ProsusAI for the financial sentiment analysis model
• yfinance: Yahoo Finance data access
• Binance API: Cryptocurrency market data
• Technical Analysis Library: Technical indicators implementation
• Open Source Community: All contributors and maintainers

📞 Support

For support and questions:

• Issues: Create an issue on GitHub
• Discussions: Use GitHub Discussions
• Documentation: Check the wiki and examples
• Community: Join our community forum

---

Happy Trading and Learning! 🚀📈

Remember: The best investment you can make is in your own education and understanding of the markets.