Title: Man vs. machine learning to time markets: Who will win?
Authors: Go Charles-Cadogan - University of Leicester (United Kingdom) [presenting]
Abstract: A market timing classifier (MTC) for high-dimensional data is introduced, which is based on a theory of portfolio manager market timing behaviour. This lies in stark contrast to the growing literature on the use of machine learning classifiers for data mining and market timing with big data in finance. The MTC separates statistically significant lower dimensional market timing portfolios from those with less timing ability. So, it falls in the class of dimension-reducing Neyman-Pearson classifiers. We applied the MTC to two different samples of high-dimensional asset pricing data, and the results show that the algorithm is able to separate statistically significant lower-dimension market timers from non-market timers. Fama-French size and value portfolios are included among significant market timers, as well as other portfolios predicated on anomalies. The MTC can be used as a pretest estimator in LASSO and machine learning classification schemes.