On data-driven robust portfolio optimization with semi mean absolute deviation via support vector clustering

In 14 the authors have studied robust semi-mean absolute deviation portfolio optimization model when assets expected returns involve uncertainty. They applied a data driven approach via support vector clustering to construct the uncertainty set using support vector clustering. In this paper, we show that their robust formulation is not the worst case counterpart of the original model. Then we give the true robust model of the underlying problems in the best an worst cases. Experiments are conducted to show the optimal objective value of the robust model in 14 belongs to the interval generated by our best and worst case models.