GRU Dynamic Beta

Estimate how a stock’s market sensitivity (beta) changes over time, powered by neural networks with built-in safeguards against lookahead bias.

from grubeta import estimate_beta

result = estimate_beta("AAPL", "SPY")
print(result["summary"])

Key Features

• Simple convenience API - Estimate beta in 3 lines of code with estimate_beta()

• Named presets - "default", "responsive", "smooth", "research" — no ML tuning needed

• GRU-based estimation - Captures complex temporal patterns in beta dynamics

• Walk-forward validation - Prevents lookahead bias with proper out-of-sample testing

• Composite loss function - Balances prediction accuracy, beta stability, alpha sparsity, and alpha temporal smoothness

• Built-in evaluation - Comprehensive metrics and benchmark comparisons

• CLI support - python -m grubeta AAPL SPY from the command line

• Production-ready - GPU support, model persistence, and extensive documentation

Quick Start

Install the package:

pip install grubeta[data]   # includes yfinance for ticker support

Estimate dynamic beta:

from grubeta import estimate_beta

# Estimate AAPL's time-varying beta relative to S&P 500
result = estimate_beta("AAPL", "SPY")
print(result["summary"])

# Access the raw beta series
result["beta"].tail()

Or use presets for different use cases:

# For event studies (captures rapid changes)
result = estimate_beta("AAPL", "SPY", preset="responsive")

# For long-term portfolio construction
result = estimate_beta("AAPL", "SPY", preset="smooth")

For advanced usage with full control:

from grubeta import DynamicBeta, DynamicBetaConfig

model = DynamicBeta(config=DynamicBetaConfig(lookback=60, lambda_beta=0.05))
results = model.fit_predict(stock_returns, market_returns, dates=dates)

Documentation

Getting Started

• Installation
  • Requirements
  • Installing from PyPI
  • Data Installation
  • Full Installation
  • Development Installation
  • GPU Support
  • Verifying Installation
  • Troubleshooting
• Quickstart
  • Estimate Beta in 3 Lines
  • Using Presets
  • Comparing Multiple Stocks
  • Using Your Own Data
  • Command Line
  • Advanced Usage
  • Understanding the Output
  • Key Parameters
  • Comparing with Benchmarks
  • Saving and Loading Models
  • Next Steps

User Guide

• Basic Usage
  • The DynamicBeta Class
  • Input Data Formats
  • The fit_predict Method
  • Separate Fit and Predict
  • Understanding Results
  • Visualization
  • Model Persistence
  • Error Handling
• Advanced Features
  • Using OHLCV Data
  • Macroeconomic Features
  • Custom Feature Engineering
  • Walk-Forward Configuration
  • Loss Function Tuning
  • Network Architecture
  • Multi-Asset Analysis
  • GPU Acceleration
  • Batch Processing
• Data Preprocessing
  • Overview
  • Lookahead Bias Prevention
  • Input Data Requirements
  • Feature Groups
  • Custom Preprocessing
  • Saving Processed Data
  • Data Quality Checks
• Model Evaluation
  • BetaEvaluator
  • Key Metrics
  • Benchmark Comparison
  • Diagnostic Tests
  • Visualization
  • Generating Reports
• Best Practices
  • Data Requirements
  • Parameter Selection
  • Avoiding Common Pitfalls
  • Production Deployment
  • Performance Optimization
  • Validation Checklist

API Reference

• Convenience API
  • format_summary()
  • estimate_beta()
  • compare_betas()
  • quick_report()
• Presets
• Exceptions
• Core API
  • DynamicBeta
  • DynamicBetaConfig
  • Example Usage
• Preprocessing API
  • DataPreprocessor
  • FeatureConfig
  • Functions
  • Example Usage
• Models API
  • GRUBetaModel
  • DualPathwayGRU
  • CAPMLoss
  • TFEnvironment
  • Architecture Overview
  • Example: Custom Model
• Evaluation API
  • BetaEvaluator
  • DiagnosticTests
  • Benchmark Functions
  • Example Usage
• Utilities API
  • Validation Functions
  • Data Manipulation
  • Performance Metrics
  • Portfolio Functions
  • Example Usage

Development

• Changelog
• Contributing

Indices and Tables

• Index

• Module Index

• Search Page

Citation

If you use GRU Dynamic Beta in your research, please cite:

@software{grubeta2026,
  author = {Yılmaz, Ahmet Selim},
  title = {GRU Dynamic Beta: Neural Network-Based Time-Varying Systematic Risk Estimation},
  year = {2026},
  url = {https://github.com/aslmylmz/grubeta}
}

License

GRU Dynamic Beta is released under the MIT License. See the LICENSE file for details.