Association Rules and Machine Learning for Enhancing Undeclared Work Detection

Abstract

Undeclared work is, by definition, a multi-faceted phenomenon that needs to be detected. In welfare states, undeclared work results in loss of public revenue and thus resources critical for welfare mechanisms’ funding, lack of worker protection and, last but not least, unfair competition for legitimate businesses. Yet, little to no studies have proposed the use of sophisticated machine learning methods in tackling this severe socioeconomic problem. In this study, we demonstrate the application of an advanced data analysis method, the association rule mining, which has significant advantages over rule-based systems, in classifying employers likely to engage in undeclared work. Indeed, the results of this pilot project proved divulging, even to the most experienced labour inspectors, offering insights in patterns of employers’ illegal behaviour, that were previously unidentified.