Leveraging Machine Learning for Official Statistics: Methodology and Application
Marco Puts* 1, Piet Daas* 1
Abstract
In this presentation, we delve into the intriguing realm of applying machine learning within the official statistical framework. Join us as we explore a comprehensive case study centered around the detection of online platforms through the analysis of website text data. We will not only unveil key findings from our research but also shed light on how this innovative application has profoundly influenced the landscape of machine learning at Statistics Netherlands.
Our primary focus revolves around addressing the critical issue of bias detection and correction in machine learning-based estimates. Through real-world examples and insights, we will illustrate how our efforts are forging a path towards a more statistical methodology in machine learning that is not only more accurate, but also fair and dependable. We will guide you through this thought-provoking journey at the intersection of data science and official statistics.