Distributionally robust sparse portfolio optimization model under satisfaction criterion

Abstract

We propose a distributionally robust portfolio optimization model with cardinality constraints under the satisfaction criterion. We aim to maximize the probability of achieving the target return of the proposed portfolio selection model while the number of assets the investors hold is limited. For practical significance, we cite a measure of shortfall-aware aspiration level to the portfolio optimization problem and convert it into a CVaR measure. In our model, we consider a worst-case and assume the distribution of returns of assets is ambiguous. We reformulate the CVaR-based measure equivalently to semi-definite programming for its tractability. A Benders' decomposition algorithm is designed to solve the proposed model efficiently. Numerical tests are utilized through actual market data to validate the proposed method. The results indicate that our algorithm can effectively solve the proposed model, and the sparse portfolio selection model under the satisfaction criterion achieves high robustness and perform better than classical models. Furthermore, we prove that taking the number of assets as the decision variable is a much more efficient method.