Empirical Asset Pricing via Machine Learning: Evidence from the European Stock Market

Abstract

This paper evaluates the performance of machine learning methods in forecasting stock returns. Compared to a linear benchmark model, interactions and non-linear effects help improve predictive performance. But machine learning models must be adequately trained and tuned to overcome the high dimensionality issue and to avoid over-fitting. Across all machine learning methods, the most important predictors are based on price trends and fundamental signals from valuation ratios. However, the models exhibit disparities in statistical performance that translate into pronounced differences in economic profitability. The return and risk measures of long-only trading strategies indicate that machine learning models produce size-able gains relative to our benchmark. Neural networks perform best, even after adjusting for risk and accounting for transaction costs. However, a classification-based portfolio formation, utilizing a support vector machine that avoids estimating stock-level expected returns, performs even better than the neural network architecture.