The US Bureau of Labor Statistics is undertaking a number of initiatives to improve the way it manages its data and metadata systems. Two examples include planning for the replacement of its public facing LABSTAT data query system and efforts within its Office of Productivity and Technology to combine multiple production systems within a single cross-divisional database platform. Within these projects, BLS views time series data as a combination of three elemental components. These components are found in all time series. They include a measure element; a person, place, and thing element; and a time element. The authors turned this basic approach into a more formal conceptual model represented in UML. The UML model describes multi-dimensional data, of which time series are a kind, and is very flexible in that it supports any kind of query into the data. The Office of Productivity has adopted the model, and it is guiding their approach moving forward. The model was also adopted by the Financial Industry Business Ontology project under the Object Management Group and, more importantly, by the DDI-4 Core development team for inclusion in that specification. There are other similarities between the OPT effort and DDI-4 Core as well. In this way, the OPT project demonstrates the feasibility and usefulness of many of the ideas in DDI-4 Core. In this talk we describe the time series formulation and the UML conceptual model. Then, the design of the OPT system and some of its features are described, relating those that are similar to DDI-4 Core where appropriate. In doing so, we provide a thorough understanding of the structure of time series, and we describe some of the productivity measures BLS/OPT produces as illustrations.

For an audio presentation coordinated with slides describing the poster see: https://doi.org/10.5281/zenodo.6620967