Unclear benefits of multi-period ML or stochastic programming approaches

Determine the extent to which adopting multi-period portfolio optimization strategies based on (i) Machine Learning–based forecasting of stock returns and (ii) stochastic programming scenario modeling yields meaningful improvements in performance or robustness compared to the single-period deterministic models studied, thereby assessing whether the increased methodological sophistication provides substantial gains.

Background

The authors propose natural extensions beyond their single-period deterministic models: a multi-period approach that either forecasts future returns using Machine Learning or models uncertainty via stochastic programming with scenarios. They question whether the increased complexity of these methods will translate into practical benefits.

This uncertainty is explicitly noted as an open consideration regarding the value of substantially more sophisticated optimization approaches, motivating an investigation into their actual performance gains.