Machine learning models predicting returns: Why most popular performance metrics are misleading and proposal for an efficient metric

作者：

Highlights：

• Many machine learning algorithms try to predict asset returns.

• Common performance metrics used to compare these algorithms are tested.

• Some metrics are misleading, most are not efficient enough.

• A more robust metric to assess the efficiency of the algorithm is proposed.

摘要

•Many machine learning algorithms try to predict asset returns.•Common performance metrics used to compare these algorithms are tested.•Some metrics are misleading, most are not efficient enough.•A more robust metric to assess the efficiency of the algorithm is proposed.

论文关键词：Stock return predictability,Machine-learning,Deep learning,Time series forecasting,Performance evaluation criteria,Investment efficiency

论文评审过程：Received 24 September 2021, Revised 9 January 2022, Accepted 21 March 2022, Available online 24 March 2022, Version of Record 28 March 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116970