Machine Learning versus the Comparables Method: Predicting Transaction Multiples in European Mid-Market M&A

Details

Supervisors

Betti, Nathanaël

Faculty

Louvain School of Management

Degree label

Master [120] en sciences de gestion, à finalité spécialisée

Abstract

enThis thesis asks whether machine learning models trained on observable company fundamentals can outperform the comparables method practitioners use to price private M&A transactions. Using a sample of 513 European mid-market Business Services transactions closed between 2015 and 2026, it compares four approaches to predicting EV/EBITDA and EV/Revenue multiples: an analyst benchmark built as the sector-year comparable-set median adjusted by a calibrated private company discount, ordinary least squares and Ridge regression as a linear control, Random Forest and XGBoost estimated on company fundamentals alone, and the same two algorithms augmented with market data.