Mortgage Rate Prediction

A machine learning project to predict 30-year fixed mortgage rates using economic indicators and Federal Reserve data.

Overview

This project uses various economic indicators and Federal Reserve data to predict future mortgage rates. It implements a Bayesian Ridge Regression model to capture uncertainties in predictions and relationships between different economic factors.

• Mortgage Rate Prediction
  • Overview
  • Sample Results (April 2025)
  • Features
  • Project Structure
  • Setup
  • Usage
    • Running the Prediction Pipeline
    • Automated Weekly Predictions
  • Model Details
  • Model Selection and Performance Analysis
    • Why These Models Performed Best
    • Feature Importance
  • Contributing
  • License

Sample Results (April 2025)

Model performance on test data:

Model	RMSE	MAE	R² Score
Bayesian Ridge Regression	0.0707	0.0468	0.6664
Linear Regression	0.0705	0.0469	0.6689

Cross-validation results:

• Bayesian Ridge Regression - Mean RMSE: 0.1066 (±0.0764)
• Linear Regression - Mean RMSE: 0.1080 (±0.0808)

Features

• Data collection from Federal Reserve Economic Data (FRED)
• Automated weekly predictions
• Time series analysis and visualization
• Weekly GitHub releases with prediction reports
• Economic indicators tracked:
  • 10-Year Treasury Yield
  • Federal Funds Rate
  • Consumer Price Index (CPI)
  • Unemployment Rate
  • GDP
  • Housing Market Indicators
  • Money Supply Metrics
  • And more...

Project Structure

mortgage-rate-prediction/
├── data/                # Data storage
├── models/             # Saved models
├── results/            # Results and visualizations
│   └── plots/         # Generated plots
├── src/               # Source code
│   ├── __init__.py
│   ├── config.py      # Configuration settings
│   ├── data_processor.py # Data collection and processing
│   ├── model_trainer.py # Model training logic
│   ├── model_evaluator.py # Model evaluation tools
│   └── utils.py       # Utility functions
└── scripts/           # Scripts for running pipelines
    ├── run_mortgage_prediction.py
    └── generate_email_report.py

Setup

1. Clone the repository:

git clone https://github.com/yourusername/mortgage-rate-prediction.git
cd mortgage-rate-prediction

2. Create a virtual environment and install dependencies:

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

3. Set up your FRED API key:
   • Get an API key from FRED
   • Add it to your environment variables: FRED_API_KEY=your_key_here

Usage

Running the Prediction Pipeline

python scripts/run_mortgage_prediction.py

Automated Weekly Predictions

The project includes a GitHub Actions workflow that:

• Runs every Monday at 00:00 UTC
• Collects latest economic data
• Generates new predictions
• Creates a GitHub release with:
  • Weekly prediction report (HTML)
  • Updated evaluation metrics
  • Latest visualization plots
• All reports are accessible through the repository's Releases page

Model Details

The project uses a Bayesian Ridge Regression model which:

• Captures uncertainty in predictions
• Handles missing data and varying scales
• Provides probabilistic predictions
• Updates beliefs based on new data

Model Selection and Performance Analysis

After evaluating 19 different regression models using PyCaret, Bayesian Ridge Regression and Linear Regression emerged as the top performers:

Why These Models Performed Best

1. Data Characteristics

   • The relationship between economic indicators and mortgage rates appears to be largely linear
   • The features show moderate multicollinearity, which Bayesian Ridge handles well
   • Weekly rate changes tend to be small and follow normal distribution patterns

2. Bayesian Ridge Advantages

   • Automatically determines optimal regularization
   • Robust to multicollinearity in economic indicators
   • Provides uncertainty estimates in predictions
   • Excellent performance with small training time (0.02 seconds)

3. Model Comparison Insights

   • More complex models (Random Forest, XGBoost) showed worse performance (R² < 0.20)
   • Tree-based models struggled, suggesting the relationship isn't highly nonlinear
   • Simple linear models outperformed ensemble methods
   • Regularized models (Ridge, Huber) performed well but not better than Bayesian Ridge

Feature Importance

The Bayesian Ridge model identified several key economic indicators that influence mortgage rate changes:

Top Positive Influencers:

• MORTGAGE30US_2 (2-month lagged mortgage rate)
• CPILFESL (Core Consumer Price Index)
• FEDFUNDS (Federal Funds Rate)
• rolling_mean_3 (3-month rolling average)
• DGS10 (10-Year Treasury Yield)
• MORTGAGE30US_4 (4-month lagged mortgage rate)
• M2SL (M2 Money Supply)

Top Negative Influencers:

• MORTGAGE30US (Current mortgage rate)
• year
• GDP (Gross Domestic Product)
• DFF (Effective Federal Funds Rate)
• GNPCA (Real Gross National Product)

Contributing

1. Fork the repository
2. Create your feature branch: git checkout -b feature/new-feature
3. Commit your changes: git commit -am 'Add new feature'
4. Push to the branch: git push origin feature/new-feature
5. Submit a pull request

License

This project is licensed under a modified GNU GPL v3 license for non-commercial use only.
See the LICENSE file for full terms.