Predictive analytics platform for CVE (Common Vulnerabilities and Exposures) publications using machine learning and statistical models. Updated daily with automated accuracy tracking.

Live Dashboard: cveforecast.org

• 13+ Forecasting Models — Statistical (Prophet, ARIMA, TBATS), ML (XGBoost, LightGBM, CatBoost), deep learning (TCN, NBEATS), and baselines
• Daily Automated Updates — GitHub Actions pipeline generates fresh forecasts at midnight UTC
• 120+ CNA Forecasts — Individual predictions for CVE Numbering Authorities with per-organization model selection
• Real-World Validation — Historical backtest on current year data with forecast vs. actual comparisons
• Monthly Self-Tuning — Automated hyperparameter optimization on the 1st of each month
• Forecast Tracking — Historical snapshots track prediction evolution and accuracy over time
• Accessible Dashboard — WCAG AA compliant with dark mode, responsive design, and keyboard navigation

• Python 3.10+
• 8GB+ RAM recommended

git clone https://github.com/RogoLabs/CVEForecast.git
cd CVEForecast
pip install -r requirements.txt

# Clone CVE data (required)
git clone --depth 1 https://github.com/CVEProject/cvelistV5.git

# Run full pipeline
python code/run_production_forecast.py

python -m http.server 8000 --directory web
# Open http://localhost:8000

code/
├── core/                  # Base classes and shared utilities
│   ├── base_forecaster.py # Abstract forecaster interface
│   ├── model_utils.py     # Shared parameter fixing and model creation
│   ├── validation_mixin.py# Cross-validation and diagnostics
│   └── data_adapter.py    # Data loading interface
├── adapters/              # Domain-specific implementations
│   ├── cve_adapter.py     # Total CVE forecasting
│   └── cna_adapter.py     # Per-CNA forecasting
├── validation/            # Statistical validation suite
├── diagnostics/           # Residual and horizon analysis
├── tuner/                 # Hyperparameter optimization
├── scripts/               # CI/CD helper scripts
└── run_production_forecast.py  # Main entry point

web/                       # Dashboard (GitHub Pages)
tests/                     # Test suite
docs/                      # Documentation

Real-world backtest accuracy on current year data:

Full rankings available on the live dashboard.

python -m pytest tests/ -v

pip install ruff
ruff check code/ tests/
ruff format code/ tests/

Models and hyperparameters are configured in code/config.json. CNA-specific settings are in code/cna_config.json.

• Architecture Guide
• API Reference
• Deployment Guide
• Development Guide
• Tuning Guide

MIT License. See LICENSE for details.

• CVE Project for the vulnerability data
• Darts for the time series forecasting framework

Version: 0.11 "Galway" 🇮🇪 | Website: cveforecast.org