12:00 - 13:00
Contributed Paper Session
Room: Upper Foyer
Chair:
Dario BUONO
Can new fairly big data sources and modernised statistical production enhance the availability of evidence based policy agenda?
Machine learning methods within the Federal Statistical Office of Germany
Florian Dumpert 1, Lydia Spies 2
1 University of Bayreuth, Bayreuth
2 Federal Statistical Office of Germany, Wiesbaden

In 2017, the Federal Statistical Office of Germany developed a Digital Agenda in order to advance the holistic digitalization of the organization. One of its main objectives is the automation of processes. First concrete classification tasks in earnings and craft statistics and in the business register have already been done. Currently, the Federal Statistical Office is working on a comprehensive integration of artificial intelligence and machine learning methods into the statistical production processes.

In the beginning, the focus will be on statistical data editing and imputation. Since those process steps are still often performed manually there is a high potential to improve efficiency through the application of machine learning techniques. The potentials and limitations of different machine learning methods for editing and imputation are therefore currently evaluated.

The presentation will give an overview of the evaluation setup and previous results as well as an outlook especially with regard to a contemplated integration of machine learning into the statistical production processes.


Reference:
Th-CPS02-02
Session:
Big Data and Machine Learning
Presenter/s:
Florian Dumpert
Presentation type:
Oral presentation
Room:
Upper Foyer
Chair:
Dario BUONO
Date:
Thursday, 18 October 2018
Time:
12:15 - 12:30
Session times:
12:00 - 13:00