A portfolio optimization model for minimizing soft margin-based generalization bound

Roy’s safety first (rsf) criterion aims to minimize the shortfall probability in portfolio selection. smoothed safety first portfolio optimization model is a useful tool to realize rsf criterion by minimizing an approximation of the empirical shortfall probability. however, the generalization performance of the smoothed safety first portfolio optimization model be poor when the number of the samples is finite. in this paper, a soft margin-based generalization bound on the shortfall probability is obtained firstly. then, a portfolio optimization model is built by minimizing the soft margin-based generalization bound. finally, the good generalization performance of the portfolio optimization model is verified by experiments.