考虑位置效应的电商平台占线品类优化问题

Online assortment optimization problem with position effect for e-commerce platforms

基于电商平台的动态特征,本文研究了一个新的品类优化问题.假设消费者行为服从多项logit(multinominal logit)选择模型,当消费者陆续到达时,平台在满足产品有限库存约束的前提下,从给定产品集合中选择一个子集,并决策对应的展示位置,提供给消费者,以最大化预期收益.以往研究建立的模型一般假设消费者类型到达序列已知或满足某随机分布,但现实中消费者到达序列往往具有高度不确定性且不可随机观测.本文利用占线理论与竞争分析的框架,在消费者类型未来到达序列未知的条件下,以最大化电商平台收益为目标,构建了同时考虑库存约束、品类优化和位置效应的占线决策模型,给出了相应的展示策略.该策略易于求解,能够快速的得到展示方案.从理论上证明了策略的竞争比并分析了问题的竞争比上界,结果说明该策略具有非常好的竞争性能.

Based on the dynamic characteristics of E-commerce platforms,this paper studies a new assortment optimization problem.Assuming that consumers'choice follows the multinominal logit model,when heterogeneous consumers arrive successively,with the goal to maximize the expected revenue,the platform needs to select a subset from the given product set and decide the display positions of products to provide to consumers,on the condition that the inventory constraints of products are satisfied.The models established by previous studies generally assumed that the arrival sequence of consumer types was known or followed some random distribution,however,in practice,the arrival sequence of consumer types is often highly uncertain and cannot be observed randomly.Under the condition that the future arrival sequence of consumer types is completely unknown,this paper constructs an online display model that considers limited inventory,assortment optimization and location effect simultaneously to maximize the revenue of the e-commerce platform,and develops a corresponding display strategy based on the online theory and competitive analysis framework.This strategy is easy to solve and an effective display scheme could be quickly obtained.The competitive performance ratio of the strategy is proved and the upper bound of the competitive performance ratio of the problem is analyzed theoretically.The results show that the strategy realizes a superior competitive performance.