Confidence-Aware Multi-Field Model Calibration
This work addresses the need for accurate probability predictions in advertisement ranking and bidding, which is critical for optimizing ad performance, but it is incremental as it builds on existing field-aware calibration methods by adding confidence-aware adjustments.
The paper tackles the problem of model calibration for predicting user feedback probabilities in advertising, where data sparsity and distribution shifts cause mismatches between predictions and true likelihoods. The proposed confidence-aware multi-field calibration method adaptively adjusts calibration intensity based on confidence levels and uses multiple fields jointly, resulting in improved advertising performance and reduced prediction deviations as shown in offline and online experiments.
Accurately predicting the probabilities of user feedback, such as clicks and conversions, is critical for advertisement ranking and bidding. However, there often exist unwanted mismatches between predicted probabilities and true likelihoods due to the rapid shift of data distributions and intrinsic model biases. Calibration aims to address this issue by post-processing model predictions, and field-aware calibration can adjust model output on different feature field values to satisfy fine-grained advertising demands. Unfortunately, the observed samples corresponding to certain field values can be seriously limited to make confident calibrations, which may yield bias amplification and online disturbance. In this paper, we propose a confidence-aware multi-field calibration method, which adaptively adjusts the calibration intensity based on confidence levels derived from sample statistics. It also utilizes multiple fields for joint model calibration according to their importance to mitigate the impact of data sparsity on a single field. Extensive offline and online experiments show the superiority of our method in boosting advertising performance and reducing prediction deviations.