Fig. 2
From: The Zoltar forecast archive, a tool to standardize and store interdisciplinary prediction research
Overview of the Zoltar Forecast Repository structure. The repository is made up of projects. Project owners define a template for a forecast by specifying units and forecast targets. They also specify the times for which forecasts should be made. They can optionally specify a “Truth Dataset”, containing eventually observed values for each target. Zoltar users designated as “Model owners” for the project can upload forecasts associated with their model via a web user interface or RESTful API (see Methods). These forecasts are run through data validation checks before entering the system. If a truth dataset is present, Zoltar generates data visualizations and scores the forecasts.
