本文提出了离散均值-方差投资组合模型的一种新的精确算法.该算法是一个基于拉格朗日松弛和Bundle对偶搜索的分枝定界算法.我们分别用随机产生的数据和美国股票市场的真实数据进行了数值实验,并与传统次梯度对偶搜索进行了比较,数值结果表明本文提出的算法对解决中小规模的离散投资组合问题是有效的.
In this paper, we propose an exact algorithm for the discrete mean-variance portfolio selection model. The algorithm is of branch-and-bound method based on Lagrangian relaxation and Bundle dual search method. Numerical experiment is carried out for test problems with data from randomly generated and U.S. stock market. Comparison results with subgradient dual search method is also reported. Computational results show that the propsed method is efficient for solving small-to-medium scale discrete mean-variance portfolio selection problems.