17.04.2012 Bank of Italy CM

From D4Science Wiki
Jump to: navigation, search

E.VanIngen, S.Bogazzi and F.Calderini attended a meeting at the Bank of Italy. The subject was their statistical model, called Matrix, more specifically for exploring possible collaborations in the area of statistical code list management, where FI is running an initiative, led by Erik.

BoI has a mature statistical model and an underpinning IS infrastructure. At technical level, they are definitively a well developed organisation. They also show a very collaborative attitude.

Related links

Summary We visited Banca D'Italia for a detailed discussion on their statistical metamodel referred to as the Matrix Model. This metamodel is at the core of their statistical systems and is currently being evaluated by Erik Van Ingen as a potential metamodel to adopt for the FI statistical Code List Manager project - a metadata management tool. During the meeting, we analyzed the structure of the metamodel, specifically in the context of the FAO Country classification usecases and some usecases related to Fisheries and fish species. BoI also expressed an interest in Master Data Management, as they also recognize that the management of statistical reference data would logically fall in the broader area of enterprise MDM. BoI expressed their interest in the Geospatial feature in the current FI statistical Code List Manager project.

Strengths of the Matrix Model

  • Temporal support - an element in a set (i.e. Turkey element in the subset European Union) could have a time period for it's membership in that subset.
  • Metamodel is expressed in Xml XSD and the Java API to reflect it is autogenerated from the XSD.
  • Role-based authorization model mapped on the metamodel concepts for data governance
  • Has matured for 20 years and goes far beyond the model of SDMX

Weaknesses/Concerns of the Matrix Model:

  • Hierarchical relationships between Domain types would have to be replicated when different domains share the same hierarchical definition.
  • It wasn’t clear if multiple code lists map to multiple Domains or Multiple Subsets.
  • It wasn’t clear if the metamodel could be used to represent graph oriented relationships
  • The metamodel is statistical-code-centric and does not start from a description of the reference entities in terms of attributes (like in a MDM approach), this can be dynamically built through the definition of ‘cubes’ (or ‘functions’)

Matrix Model Illustration: MatrixBoI120318.jpg