Abstract

With a growing demand for more timely, accurate, relevant and trustworthy statistics to inform society and decision makers, National Statistics Organisations (NSOs) have been investigating innovative ways to produce official statistics more efficiently and enhance their data ecosystems. The Machine Learning (ML) Group is led by the Office for National Statistics (ONS) in partnership with the UN Economic Commission for Europe High-Level Group for the Modernisation of Official Statistics. It provides a platform for the global statistical community to develop joint research, build capability and share knowledge on ML developments in official statistics.
This talk will cover the quality framework for statistical algorithms, which explores the development of a quality framework to compare different ML methods based on a standard set of five criteria: i) explainability; ii) accuracy; iii) reproducibility; iv) timeliness and v) cost-effectiveness. The Group has also advanced on new areas, such as: acquisition of high-quality training data sets, model monitoring and user interfaces development. As privacy and ethical concerns grow and public awareness of artificial intelligence (AI) increases, NSOs will need to establish robust ecosystems to address these issues, e.g.: responsible machine learning.