Gloria is a modern open-source framework for time series analysis and forecasting, designed for the demands of complex, real-world data. It combines robust statistical modeling with flexible controls and full transparency to enable trustworthy forecasting.
- Distributional Flexibility: Go beyond the normal distribution and model count data (Poisson, Binomial, Negative Binomial, Beta-Binomial), bounded rates (Beta), or non-negative floats (Gamma) natively
- Any Time Grid: Gloria handles arbitrary sampling intervals (not just daily)
- Rich Event Modeling: Parametric and extensible event library to handle holidays, campaigns, or maintenance windows - any event, any shape, for realistic impacts and reduced overfitting.
- Fully Explainable: Gloria's models are explicit, fully documented, and always inspectable.
- Modern Python Stack: Type hints, pydantic for validation, and a clean API design reminiscent of Prophet but with a much more maintainable and extensible codebase.
- Documentation: https://e-dyn.github.io/gloria/
- Installation: https://e-dyn.github.io/gloria/get_started/installation.html
- Source Code: https://github.com/e-dyn/gloria
- Bug Reports and Feature Requests: https://github.com/e-dyn/gloria/issues
- License: MIT