64th ISI World Statistics Congress - Ottawa, Canada

64th ISI World Statistics Congress - Ottawa, Canada

Have innovative machine learning techniques boosted official statistical production in Switzerland?


64th ISI World Statistics Congress - Ottawa, Canada

Format: CPS Abstract

Session: CPS 26 - Official statistics II

Monday 17 July 5:30 p.m. - 6:30 p.m. (Canada/Eastern)


A few years ago, the Swiss Federal Statistical Office (FSO) adopted a data innovation strategy that included the launch of innovation pilot projects. The purpose of this paper is to evaluate the situation five years after the launch of these projects to draw practical lessons from them and to know the impact on the statistical production at FSO.

After having recalled the criteria developed to choose and finally retain which innovation pilot projects to pursue, we propose to describe these projects by focusing on their expected added value, the problems encountered and the solutions provided.
Note that the projects mentioned here all implement techniques from artificial intelligence such as machine learning or deep learning. They thus cover a wide range of examples ranging from aerial image recognition or plausibility check to automatic text recognition.
An example will serve as a guideline for the delicate transition from the experimental stage to the implementation in statistical production. In this context, the criteria for approving a transfer to statistical production will be discussed and commented in order to be able to objectively answer the question asked in the title!