基于重取样的鲁棒投资组合优化的加速解法

Acceleration solving method for robust portfolio optimization based on resampling

针对基于重取样的鲁棒投资组合优化模型,为有效解决重取样和混合整数规划模型计算量太大的问题,提出一种高效的加速求解方法。利用带初始解和不带初始解的双线程来求解问题,始终选取计算快的线程,加速整个求解过程,初始解的寻找依赖已经求解的模型结果和模型的鲁棒性特征。数值实验选取模拟数据对该加速方法及其性能进行分析,实验结果表明,在几乎不影响最优值的情况下,该方法使计算效率得到较大提高。

To accelerate the solving process of the robust portfolio optimization model based on resampling,an efficient method was presented.The process was accelerated by choosing the faster thread from two threads,in which one started with a primal solution and the other did not.The fact that the primal feasible solution depended on the results of the solved models and the ro-bustness of the model was found.This experiment made analysis of the performance of the proposed method with simulation da-ta.Experimental results show that this method can save considerable calculating time when the optimal value is achieved.