通过多选背包缩减实现稳健离散定价优化
A Certifying MCKP Framework for $Γ$-Robust Discrete Pricing
AI总结:
本文研究了在边际和公平性约束下选择每个产品的单一价格的离散投资组合定价问题,通过将问题转化为多选背包问题,揭示了线性规划松弛的结构特性,并展示了固定参数子问题的整数规划间隙有界性。
摘要:
我们研究了一个离散投资组合定价问题,该问题在边际和公平性约束下为每个产品选择一个价格。为应对需求不确定性,我们引入了一个预算鲁棒公式,以控制保守性同时保持计算可行性。通过将问题转化为多选背包问题(MCKP),我们识别了线性规划松弛的结构特性,特别是上包过滤和 hull 段上的贪心填充,这些特性为固定参数子问题的线性规划松弛提供了精确解法。对于由此产生的固定参数子问题,我们证明了整数规划间隙有界于单个项 hull 跳跃,并且在标准有界性和线性增长假设下,相应的相对间隙以 O(1/n) 的速率衰减。在合成投资组合和一个 stylized 零售案例研究中,经济校准的参数与这些界限一致,并表明在测试实例中,鲁棒边际保护可以通过小于1%的名义收入损失实现。
英文摘要:
We study finite-menu portfolio pricing under a ratio margin requirement, price-admissibility constraints, and integer-budget demand uncertainty. The pricing model reduces exactly to a multiple-choice knapsack problem (MCKP), while the coupled robust constraint is equivalent to a finite family of fixed-threshold MCKPs indexed by zero and every original option deviation. For each threshold, classical upper-hull geometry yields an LP bound and a feasible one-item rounding rule whose loss is bounded by a single adjacent hull-value jump; under bounded jumps and linear objective growth, the corresponding relative bound is $O(1/n)$. An exact-arithmetic branch-and-bound construction provides either a global optimum or a valid anytime gap over the complete threshold family, and direct evaluation of the original sorted-$Γ$ protection term checks every returned pricing decision. The contribution is an end-to-end, pricing-specific certifying framework that connects these ingredients, states their boundary conditions, and implements the full original breakpoint family without tolerance clustering.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。