T-Learner Based on Machine Learning for Customer Conversion Prediction and Causal Effect Evaluation

Increasing customer conversion rate is an important topic of concern for every business entity. Among them, promotions and offers are the most common ways to do so. Often, managers want to predict whether customers are likely to become repeat customers through their consumption behaviour and quantify the impact of coupon issuance or not on customer conversion rate to help them determine practical business strategies. In this paper, we focus on customer consumption data of commercial entities, and construct a repurchase prediction model based on three different machine learning algorithms, namely Random Forest, XGBoost and LightGBM, based on RFMA consumption behaviour characteristics. Meanwhile, using the repurchase prediction model combined with the T-learn evaluation method, we propose a quantitative indicator of AUUC for evaluating the impact of coupon issuance on customer conversion rate. The machine learning-based quantitative model for interpretable causal inference proposed in this paper has feasible guidance and value in helping decision makers adopt optimal business strategies.