FreqAI - Machine Learning Integration

FreqAI is Freqtrade's machine learning integration that allows you to use various ML models to enhance your trading strategies.

Documentation Status

The FreqAI documentation is currently being converted from the original MkDocs format. The following sections will be available soon:

• Introduction - Overview of FreqAI capabilities
• Configuration - Setting up FreqAI in your bot
• Parameter Table - Complete parameter reference
• Feature Engineering - Creating and managing features
• Running FreqAI - Operating FreqAI-enabled strategies
• Reinforcement Learning - Advanced RL integration
• Developer Guide - Contributing to FreqAI development

What is FreqAI?

FreqAI is a powerful machine learning framework integrated into Freqtrade that enables:

• Predictive Modeling - Use historical data to predict future price movements
• Feature Engineering - Automatically create and manage technical indicators as features
• Model Training - Train various ML models including neural networks, random forests, and more
• Live Prediction - Make real-time predictions during trading
• Reinforcement Learning - Advanced RL agents for strategy optimization

Key Features

Supported Models

• Scikit-learn models (Random Forest, SVM, etc.)
• XGBoost and LightGBM
• Neural Networks via TensorFlow/Keras
• Reinforcement Learning agents
• Custom models through extensible framework

Data Management

• Automatic feature creation from OHLCV data
• Feature normalization and preprocessing
• Data splitting for training/validation/testing
• Rolling window training for time series data

Integration

• Seamless strategy integration - Use predictions in your strategies
• Backtesting support - Test ML strategies with historical data
• Live trading - Deploy ML models in production
• Performance monitoring - Track model performance over time

Getting Started

To use FreqAI, you'll need to:

1. Install dependencies - Additional ML libraries
2. Configure FreqAI - Add FreqAI section to your config
3. Create ML strategy - Implement FreqAI-enabled strategy
4. Train models - Run training on historical data
5. Deploy and monitor - Use in live trading

Example Configuration

{
    "freqai": {
        "enabled": true,
        "identifier": "example_freqai_strategy",
        "feature_parameters": {
            "include_timeframes": ["5m", "15m", "4h"],
            "include_corr_pairlist": ["ETH/USD", "LINK/USD", "BNB/USD"],
            "label_period_candles": 24,
            "include_shifted_candles": 2,
            "DI_threshold": 0.9,
            "weight_factor": 0.9,
            "principal_component_analysis": false,
            "use_SVM_to_remove_outliers": true,
            "stratify_training_data": 0,
            "indicator_periods_candles": [10, 20]
        },
        "data_split_parameters": {
            "test_size": 0.33,
            "shuffle": false
        },
        "model_training_parameters": {
            "n_estimators": 1000
        }
    }
}

Next Steps

Once the full FreqAI documentation is available, you'll be able to:

• Learn detailed configuration options
• Understand feature engineering techniques
• Implement custom ML models
• Optimize model performance
• Deploy production ML strategies

For now, please refer to the original Freqtrade documentation at freqtrade.io for complete FreqAI information.