Data Mining for Targeted Inspections Against Undeclared Work by Applying the CRISP-DM Methodology

Abstract

Undeclared work is a complex and explicitly hidden socio-economic problem that can take various forms and its consequences are substantial to states, employees and businesses. Thus, inspection authorities need to use their resources more effectively and cleverly to tackle this unlawful phenomenon successfully. This study presents the use of an advanced machine learning method – Associative Classification – through a data mining application following the CRISP-DM methodology, for more effective selection methods of on-site labour inspections against undeclared work. The study proves that the produced classifier can, on the one hand highly contribute in scheduling targeted inspections of increased efficiency and, on the other hand offer actionable and comprehensible knowledge to the labour inspectors regarding the employers’ illegal practices related to undeclared work and other labour law infringements.