Relative Valuation with Machine Learning

ABSTRACT We use machine learning for relative valuation and peer firm selection. In out‐of‐sample tests, our models substantially outperform traditional in accuracy. This outperformance persists over time holds across different types of firms. The valuations produced by behave like fundamental values. Overvalued stocks decrease price undervalued increase the following month. Determinants multiples identified are consistent with theoretical predictions derived from a discounted cash flow approach. Profitability ratios, growth measures, efficiency ratios most important value drivers throughout sample period. derive novel method to express predicted as weighted averages multiples. These weights measure peer–firm comparability can be used selecting peer‐groups.