OPERA: Open-Source Prediction Evaluation and Real-Time Analysis¶

OPERA is a modular ecosystem designed to support the use of time-series models in real-time settings and foster open collaboration. It comprises the following modules:

• opera-eco pins compatible releases and supplies documentation, AI skills, and integration tests.
• forecast_evaluation validates vintaged outturns and forecasts and provides evaluation and visualisation capabilities.
• forecast_realtime runs models across data vintages.
• forecast_combo combines forecasts through averaging, regression, error-based weighting, or hierarchies.
• news_decomp attributes nowcast levels and revisions to news, re-estimation, and interaction.

OPERA supports a broad range of models through wrappers for libraries such as scikit-learn and R's fable. You can also add your own model. In addition to these wrappers, OPERA includes models with native support:

• bvar provides tools for working with Bayesian VARs.
• nowcast-midas nowcasts quarterly targets from higher-frequency indicators using MIDAS and combination techniques.

Licence and copyright¶

This project is released under the MIT Licence.

Copyright © 2026 Bank of England.

The full licence text is available in the repository's LICENSE file.