Clustering Context in Off-Policy Evaluation
This work addresses a specific bottleneck in off-policy evaluation for applications like e-commerce and healthcare, offering an incremental improvement over existing methods.
The paper tackles the problem of deteriorating performance in off-policy evaluation when logging and evaluation policies differ significantly, by proposing an estimator that clusters similar contexts, and experimental results show it improves estimation accuracy, particularly in deficient information settings.
Off-policy evaluation can leverage logged data to estimate the effectiveness of new policies in e-commerce, search engines, media streaming services, or automatic diagnostic tools in healthcare. However, the performance of baseline off-policy estimators like IPS deteriorates when the logging policy significantly differs from the evaluation policy. Recent work proposes sharing information across similar actions to mitigate this problem. In this work, we propose an alternative estimator that shares information across similar contexts using clustering. We study the theoretical properties of the proposed estimator, characterizing its bias and variance under different conditions. We also compare the performance of the proposed estimator and existing approaches in various synthetic problems, as well as a real-world recommendation dataset. Our experimental results confirm that clustering contexts improves estimation accuracy, especially in deficient information settings.