因果推断下基于加速失效时间模型的网络营销策略研究

Research on Internet Marketing Strategy Based on Accelerated Failure Time Model under Causal Inference

摘要: 本文构建了基于倾向性得分的加速失效时间模型，探究电商平台促销活动与用户复购行为之间的因果关系。研究采用倾向性得分处理观测数据中的选择偏倚问题，并利用Logistic回归模型对其进行估计。随后，将倾向性得分作为协变量纳入加速失效时间模型，对促销活动与用户复购时间间隔的关系进行建模，并估计因果效应，同时考虑了右删失数据的处理。实证结果显示，在对一系列关键混杂因素进行调整后，是否使用优惠券的促销行为，对用户复购时间间隔产生统计学上的显著影响，证明了模型的有效性。一定程度上，促销活动可以加速用户复购速度，帮助电商平台调整网络营销策略。

Abstract: This article constructs an accelerated failure time model based on propensity scores to explore the causal relationship between promotional activities on e-commerce platforms and user repurchase behavior. The study uses propensity scores to address the selection bias problem in observational data and estimates it using a logistic regression model. Subsequently, the propensity score is incorporated into the accelerated failure time model as a covariate to model the relationship between promotional activities and the time interval between user repurchases, and the causal effect is estimated, while also considering the treatment of right-censored data. The empirical results show that after adjusting for a series of key confounding factors, the promotional behavior of using coupons has a statistically significant impact on the time interval between user repurchases, which proves the validity of the model. To a certain extent, promotional activities can accelerate the speed of user repurchases and help e-commerce platforms adjust their online marketing strategies.