12:30 - 13:30
Poster Session
Room: Lunches Space
A Data Integration System for RIAD Bundesbank
Katja Ziprik, (Email)
Deutsche Bundesbank, Frankfurt Main
The Register of Institutions and Affiliates Data (RIAD) is a shared register maintained by the European System of Central Banks (ESCB). Each National Central Bank of the Eurozone provides input for RIAD in its own area of competence, whereas RIAD Bundesbank is the corresponding German register. It contains the reference data on legal and other statistical institutional units and facilitates the integration of several databases, especially by allocating common identifiers, attributes and metadata. It will provide a data hub function by integrating and processing multiple internal and external databases, f. e. the Analytical Credit Database (AnaCredit) and the Deutsche Bundesbank Prudential Database. It enables a high flexibility with regards to analysis; hence the collected data can be used for statistical and non-statistical purposes, both across institutions and across different user groups within institutions. At the moment, there are over 1.700 reporting agents from the AnaCredit primary reporting and already three internal registers that are continuously integrated into the system. These data sources differ significantly in their quality and data structures. In the near future, further commercial and official data sources will be gradually integrated. There is no stable national identifier with full coverage available in Germany to describe the respective reporting units. The aim of the integration system of RIAD Bundesbank is to classify (“match”/ “non-match”) and consolidate every received reporting unit on a highly automated basis, since the data throughput is too high to deal with it manually. To fulfil the formulated requirements of the corresponding regulation (ECB/2016/13) and guideline (ECB/2018/16), the produced data quality needs to be very high. Thus the implemented algorithms are precision oriented. For the task of record linkage there are already several deterministic stages implemented. A machine learning based record linkage prototype has been developed to complement the deterministic stages. For the consolidation of the data pairs, RIAD Bundesbank contains a highly automated block-compounding algorithm which takes the veracity, velocity and modus of the reported attribute into account.


Reference:
POST01-003
Session:
Big data analytics (poster)
Presenter/s:
Katja Ziprik
Presentation type:
Poster presentation
Room:
Lunches Space
Date:
Tuesday, 12 March
Time:
12:30 - 13:30
Session times:
12:30 - 13:30