Well-calibrated predictions of user preferences are essential for many applications. Since recommender systems typically select the top-N items for users, calibration for those top-N items, rather than for all items, is important. We show that previous calibration methods result in miscalibrated predictions for the top-N items, despite their excellent calibration performance when evaluated on all items. In this work, we address the miscalibration in the top-N recommended items. We first define evaluation metrics for this objective and then propose a generic method to optimize calibration models focusing on the top-N items. It groups the top-N items by their ranks and optimizes distinct calibration models for each group with rank-dependent training weights. We verify the effectiveness of the proposed method for both explicit and implicit feedback datasets, using diverse classes of recommender models.

Calibration is defined as the ratio of the average predicted click rate to the true click rate. The optimization of calibration is essential to many online advertising recommendation systems because it directly affects the downstream bids in ads auctions and the amount of money charged to advertisers. Despite its importance, calibration optimization often suffers from a problem called "maximization bias". Maximization bias refers to the phenomenon that the maximum of predicted values overestimates the true maximum. The problem is introduced because the calibration is computed on the set selected by the prediction model itself. It persists even if unbiased predictions can be achieved on every datapoint and worsens when covariate shifts exist between the training and test sets. To mitigate this problem, we theorize the quantification of maximization bias and propose a variance-adjusting debiasing (VAD) meta-algorithm in this paper. The algorithm is efficient, robust, and practical as it is able to mitigate maximization bias problems under covariate shifts, neither incurring additional online serving costs nor compromising the ranking performance. We demonstrate the effectiveness of the proposed algorithm using a state-of-the-art recommendation neural network model on a large-scale real-world dataset.