Optimizing Risk-balancing Return under Discrete Choice Models