Shining Light into the Machine Learning Black Box
Also Interesting
While machine learning models are increasingly being used to predict returns, little progress has been made to leverage models for factor portfolio attribution, a critical component in analysing systematic portfolio returns. What are the limitations of existing methodologies, what is Shapley value, how is it applied and how can a SHAP based performance attribution framework be combined with non-linear models to interpret black box models and to better understand factor contributions?
