多产品报童问题的直接搜索算法求解  

Direct Search Method for Multi-Product News Vendor Problem

作　　者：郝爽 张大力[1] 董明[1] HAO Shuang;ZHANG Dali;DONG Ming(Antai College of Economics and Management,Shanghai Jiao Tong University,Shanghai 200030,China)

机构地区：[1]上海交通大学安泰经济与管理学院,上海200030

出　　处：《上海管理科学》2023年第5期5-8,72,共5页Shanghai Management Science

基　　金：国家自然科学基金(72071128,71931007)。

摘　　要：多产品报童问题假设商品需求服从已知分布,在实践中难以直接应用,可将零售商收益视作表达式未知的随机黑箱函数,通过样本均值对收益的期望进行近似,并利用直接搜索算法最大化零售商收益。数值实验表明,直接搜索算法结合可变数量的样本均值近似可在样本数量极为有限的条件下有效求解多产品报童问题,且求解质量优于已有的启发式算法。In this research,multi-product news vendor problem with resource constraint is modeled as a stochastic black-box optimization problem without assuming known distribution of demands.A directional direct search method with variable number samples is designed to solve it.Numerical test shows proposed method can solve multi-product news vendor problem with resource constraint with limited samples and results is better than known heuristics.

关 键 词：多产品报童问题 随机黑箱优化 直接搜索算法 可变样本数量 

分 类 号：F722.2[经济管理—产业经济]