基于局部斜率更新的数据驱动动态定价策略

基于某经济型连锁酒店的实际入住数据,研究针对未知需求的数据驱动动态定价策略的应用.提出一种新的非参数的需求函数刻画方法,称为"局部斜率更新".该方法替代以往数据驱动策略中所使用的最小二乘法,得到新的数据驱动定价策略:基于"局部斜率更新"方法的短视(被动学习)动态定价策略,称为IL策略;基于"局部斜率更新"方法的主动学习动态定价策略,称为CIL策略.新方法能够克服价格内生性的难题,解决以往数据驱动策略的应用困境.数据实验表明,新方法的拟合优度接近最小二乘法.在线性和非线性需求函数形式的假设下,"局部斜率更新"方法的拟合优度与最小二乘法分别只相差7.19%和4.69%.酒店实际数据验证表明,结合新方法的数据驱动定价策略明显优于酒店经理的实际定价决策,其中IL可降低60%的后悔值, CIL可降低77.59%的后悔值.本文提出的需求函数刻画方法及结论有助于指导未知需求数据驱动收益管理理论的应用实践.

Data-driven revenue management strategy based on local-slope update

Based on the real data of one economical hotel group,this paper studies the application of datadriven revenue management with unknown demand.To tackle price endogeneity,a new non-parametric approach of demand estimation,called local-slope update,is proposed to replace the OLS used by current datadriven policies.New dynamic pricing policies,namely iterated local-slope(IL)and constrained iterated localslope(CIL),are also developed.The results show that the forecasting accuracy of local-slope update is less than that of OLS only by 7.19%and 4.69%for linear and nonlinear demands,respectively.Both the IL policy and CIL policy can significantly improve hotel revenue.The CIL can reduce 77.59 percent of the regret value and IL can reduce 60 percent of the regret value.This non-parametric approach of demand estimation helps the applications of data-driven revenue management methods in practice.