Stock Returns Forecasting via Machine Learning with Average Windows Forecasts

Stock returns predictability has been a long-standing topic in the literature on financial economics. Developments in prediction technology have facilitated the wide use of machine learning techniques, which motivates our study of whether stock returns predictability can be improved using machine learning. Methodologically, we apply machine learning methods to revisit Chen et al. (1986) under the framework of average window forecasts (AveW) of Pesaran and Pick (2011). To account for out-of-sample model uncertainties, we conduct forecasting combination by averaging all AveWs, which verifies forecasting improvement. Finally, evidence from Diebold and Mariano test shows that both approaches perform equally well