Fairness-Aware Re-Ranking and Explanation in Top-N Recommendation

This document describes a Python software artifact prepared for the CAEPIA 2026 tutorial. The implementation combines (i) a matrix-factorization recommender trained with Cornac, (ii) a calibrated-fairness (FCal) top-K selection procedure formulated as an integer linear optimization problem and solved with PuLP, and (iii) a global decision-tree model that provides feature-based descriptions of the item-level changes introduced by re-ranking. The artifact is intended for instructional and experimental use. It is a tutorial-oriented implementation related to the methodological line presented in Behind the Re-Rank: Explaining Fairness in Top-N Recommendations (UMAP 2026), but it is a separate code implementation and should not be assumed to be identical to the software repository associated with that publication.