Mean Absolute Directional Loss as a new loss function for machine learning problems in algorithmic investment strategies

Jakub Michanków, Pawel Sakowski, Robert Slepaczuk

Abstract: Highlights•New concept of loss function more adequate in ML methods used in AIS.•Comparison of state-of-the-art loss function used in ML optimization problems.•Novel solution of the important overoptimization problem of ML models.•Increase of the efficiency of out-of-sample investment strategies results.•Solution to the problem of hyperparameters tuning.