Using AI for Predicting Cross-section of Expected Returns

Abstract

The number of cross-sectional predictors has soared in the last two decades and that made this area of research as “factor zoo”. Recent trend in asset pricing research focus on how we sort out these factors. From practitioner’s viewpoint, factor choice is the most important task for the successful asset management. We discuss the possibility of AI (machine learning and non-linear estimation) in optimizing portfolio given the past factor performance. We also discuss the return predictability of the model which uses data with limited availability.