Enhancing Counterfactual Classification via Self-Training

TL;DR

This paper introduces Counterfactual Self-Training (CST), a novel method that improves counterfactual classification by imputing outcomes for unseen actions, addressing bias in observational data through iterative pseudolabeling and input consistency loss.

Contribution

It proposes a self-training algorithm for counterfactual classification that effectively imputes outcomes for unseen actions, enhancing performance in biased observational data scenarios.

Findings

CST outperforms baseline methods on synthetic datasets.

CST achieves significant improvements on real-world datasets.

Input consistency loss further enhances CST accuracy.

Abstract

Unlike traditional supervised learning, in many settings only partial feedback is available. We may only observe outcomes for the chosen actions, but not the counterfactual outcomes associated with other alternatives. Such settings encompass a wide variety of applications including pricing, online marketing and precision medicine. A key challenge is that observational data are influenced by historical policies deployed in the system, yielding a biased data distribution. We approach this task as a domain adaptation problem and propose a self-training algorithm which imputes outcomes with categorical values for finite unseen actions in the observational data to simulate a randomized trial through pseudolabeling, which we refer to as Counterfactual Self-Training (CST). CST iteratively imputes pseudolabels and retrains the model. In addition, we show input consistency loss can further improve CST performance which is shown in recent theoretical analysis of pseudolabeling. We demonstrate the effectiveness of the proposed algorithms on both synthetic and real datasets.