RT Journal Article
SR Electronic
A1 Staňková, Veronika
T1 (Out)smart the Peer Group in Market Comparison: Building Business Valuation Multiples by Machine Learning
JF European Journal of Business Science and Technology
YR 2024
VO 10
IS 2
SP 156
OP 172
DO 10.11118/ejobsat.2024.011
UL https://ejobsat.cz/artkey/ejo-202402-0007.php
AB Traditionally, market comparison requires identifying a peer group, which still poses unresolved practical difficulties today. This research seeks to provide valuable insights into the practicality, efficiency, and accuracy of machine learning in valuing a company. It employs a state-of-the-art machine learning technique, Gradient Boosting Decision Trees (GBDT), to predict the valuation multiple directly. A yearly dataset of U.S. public companies from 1980-2021 was used. The most common multiples (EV/EBITDA, EV/EBIT, P/E, and EV/Sales) were tested. The performance of GBDT was assessed against an industry-based method. GBDT consistently outperformed the alternative method with an average 24 percentage point decrease in the median average percentage error. The results support GBDT's potential as a supplementary tool in valuation practice.