Multi-Target Learning for Empirical Asset Pricing​ (Machine Learning)

  • Empirical asset pricing focuses on the predictability of stock returns in the cross-section, m(1) = first moment.​
  • Predictors are mainly extracted from stock characteristics (see "factor zoo").​
  • Recent literature started to include option-implied characteristics ("Q") for predicting m(1).​
  • We want to consider multi target learning, i.e. learning m(1) to m(5) simultaneously while using common stock characteristics as well as P and Q information and applying numerous methodologies that have been suggested by the literature.