TSCAN: Context-Aware Uplift Modeling via Two-Stage Training for Online Merchant Business Diagnosis

2025-04-26

AI Summary

To address inaccurate individual treatment effect (ITE) estimation caused by sample selection bias and dynamic external contextual influences, this paper proposes TSCAN, a two-stage context-aware boosting modeling framework. Methodologically, TSCAN features: (1) a novel CAN-U/CAN-D two-stage architecture—generating counterfactual uplift labels in the first stage and performing explicit bias correction via an isotropic output layer in the second; (2) a context-aware attention mechanism that captures fine-grained interactions among treatment, merchant, and multidimensional environmental features; and (3) integration of IPM regularization, propensity-score guidance, and joint training. Evaluated on two real-world business datasets, TSCAN significantly outperforms state-of-the-art uplift modeling approaches. Deployed at China’s largest online food delivery platform, it supports operational decision-making for over ten million merchants, achieving a 12.7% improvement in uplift prediction accuracy in production.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Algorithm Configuration
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

A primary challenge in ITE estimation is sample selection bias. Traditional approaches utilize treatment regularization techniques such as the Integral Probability Metrics (IPM), re-weighting, and propensity score modeling to mitigate this bias. However, these regularizations may introduce undesirable information loss and limit the performance of the model. Furthermore, treatment effects vary across different external contexts, and the existing methods are insufficient in fully interacting with and utilizing these contextual features. To address these issues, we propose a Context-Aware uplift model based on the Two-Stage training approach (TSCAN), comprising CAN-U and CAN-D sub-models. In the first stage, we train an uplift model, called CAN-U, which includes the treatment regularizations of IPM and propensity score prediction, to generate a complete dataset with counterfactual uplift labels. In the second stage, we train a model named CAN-D, which utilizes an isotonic output layer to directly model uplift effects, thereby eliminating the reliance on the regularization components. CAN-D adaptively corrects the errors estimated by CAN-U through reinforcing the factual samples, while avoiding the negative impacts associated with the aforementioned regularizations. Additionally, we introduce a Context-Aware Attention Layer throughout the two-stage process to manage the interactions between treatment, merchant, and contextual features, thereby modeling the varying treatment effect in different contexts. We conduct extensive experiments on two real-world datasets to validate the effectiveness of TSCAN. Ultimately, the deployment of our model for real-world merchant diagnosis on one of China's largest online food ordering platforms validates its practical utility and impact.

Problem

• Addresses sample selection bias in ITE estimation
• Reduces information loss from traditional regularization techniques
• Models varying treatment effects across different contexts

Innovation

• Two-stage training with CAN-U and CAN-D models
• Context-Aware Attention Layer for feature interactions
• Isotonic output layer to directly model uplift effects