Pricing Analytics: Modeling, Theory and Algorithms

该奖项的目标是开发考虑到相关的、经经验验证的消费者行为的新的动态定价模型。基于来自行业合作伙伴和开放数据库的经验数据，我们计划构建准确和易处理的需求模型，以适应定价优化。具体地说，我们将探索在动态定价下纳入消费者基于记忆的参考价格的需求模型。然后我们将使用这些需求模型来构建定价优化模型，该模型考虑了各种复杂的实际操作约束。该项目的需求模型和决策模型从单一产品到多产品，从耐用产品到易腐烂产品，从确定性设置到随机设置。通过利用这些模型的特殊结构来开发先进的分析技术和高效的算法将是该项目的关键。如果成功，这项研究将导致新的和全面的分析模型，以及先进的方法和有效的算法，可以用来解决由此产生的具有挑战性的非光滑、非凸优化问题。这些模型、方法和算法对于开发急需的决策支持系统以提高公司的竞争优势至关重要。初步研究表明，基于记忆的参考价格模型导致了复杂的价格动态，并提出了一系列具有挑战性但实际的研究问题，而现有文献尚未解决这些问题。我们期望开发的新的理论和技术不仅将成功地解决这些问题，而且还将在其他研究领域如动态系统中有用。

The objective of this award is to develop novel dynamic pricing models that take into account pertinent, empirically validated consumer behaviors. Building upon empirical data from industrial partners and open databases, we plan to construct accurate and tractable demand models that are amenable to pricing optimization. Specifically, we will explore demand models that incorporate consumers' memory-based reference prices under dynamic pricing. We will then use these demand models to build pricing optimization models that take into account a variety of complex practical operation constraints. The project's demand models and decision models range from single products to multiple products, from durable products to perishable products, and from deterministic settings to stochastic settings. Developing advanced analytical techniques and efficient algorithms by exploiting the special structures of these models will be essential to this project. If successful, this research will lead both to novel and comprehensive analytical models and to advanced methodologies and efficient algorithms that may be used to attack the resulting challenging non-smooth, non-convex optimization problems. These models, methodologies, and algorithms will be critical for the development of the much needed decision support systems to improve companies' competitive advantages. Preliminary research demonstrates that memory-based reference price models result in complex price dynamics and raise a host of challenging yet practical research questions that the existing literature does not yet address. The novel theory and techniques we expect to develop will not only successfully address these questions, but will also be useful in other research areas such as dynamic systems.