All Times ET
Program is Subject to Change
Machine Learning for Classification, Imputation, and Outlier Detection in German Establishment Statistics (307938)*Florian Dumpert, Federal Statistical Office of Germany
Keywords: machine learning, official statistics, quality, efficiency
Machine learning methods, often conceptionally known since the first half of the last century, can nowadays be used due to increasing computational power and storage capacities. This does not mean that they necessarily have to be implemented into the processes in official statistics - but official statistics should have an answer whether the usage of machine learning would improve the quality of statistics, would offer additional options for analysis on official data, or would be able to make processes more efficient. We show the developments in the field of ML in German official statistics using concrete examples in establishment statistics.