The paper explores the possibility of implementing machine learning (ML) techniques to obtain an automatic classification by institutional sector of the Italian companies recorded in the Bank of Italy's 'Entities Register'.
The classifier yields accurate predictions of the sector of economic activity code for more than 99% of the corporations analysed, even taking into account the unbalanced nature of the population of interest, which mainly comprises non-financial corporations. The selected algorithm will make the entire classification process significantly more efficient by reducing the area of manual intervention.
