离散到凸性:通过简化拓扑三角化实现促销计划
Discreteness to Convexity: Promotion Planning via Simplotope Triangulation
AI总结:
本文通过简化拓扑三角化方法,提出了一种新的价格促销优化公式化方法,显著提升了求解效率,并扩展了线性规划松弛条件的应用范围。
摘要:
价格促销优化是一个计算上具有挑战性的问题,对超市运营至关重要,要求在多个产品和时期内同时做出定价决策。本文介绍了一种新的公式化方法,用于超模函数和单变量组合,通过从简化拓扑三角化中获得的显式凸包描述来构建,这与以往依赖矩形域的做法不同。利用这种重新公式化方法与Gurobi相结合,我们实现了显著的性能提升,对于具有10个产品和5个价格层级的问题,平均求解时间从434秒减少到0.06秒,使实例规模显著扩大。我们展示了紧线性规划松弛条件的条件,将先前结果从两个价格层级扩展到多个价格层级,并从加法历史效应扩展到乘法历史效应。我们的方法广泛适用于非线性离散优化,我们贡献了将任意单变量函数的组合凸化的技术,以及将L自然函数的超类凸化的框架,提供了强大的收入管理工具。这项工作提高了价格促销优化的可处理性,为大规模超市运营提供了一个实用且理论坚实的解决方案。
英文摘要:
Price promotion optimization is a computationally challenging problem central to supermarket operations, requiring simultaneous pricing decisions across multiple products and periods. This paper introduces a new formulation for price promotion by developing convex hull results for supermodular compositions of univariate functions over a simplotope. Leveraging this reformulation with Gurobi, we achieve substantial performance gains: instances with up to 125 products, 20 periods, and 5 price levels are solved in an average of 7 minutes, demonstrating the potential to handle even larger instances. Our exact solution methods extract 25--48\% additional profit from promotion planning relative to state-of-the-art heuristic approaches. Additionally, we extend the polynomially solvable cases from two to multiple price levels and expand our results to allow for multiplicative historical effects. Our core methodological innovation applies to a broad class of nonlinear discrete optimization problems. Specifically, our results convexify a class of nonlinear functions that includes monomials and the widely studied L natural function structure.
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。