Ranking Aggregation with Weak Order Outputs
We study the problem of aggregating rankings when ties are allowed in the input, leading to weak order outputs. We propose an optimization-based model that incorporates these ties and enforces consistency in the final ranking, allowing partial comparisons between items. We analyse three variants of the model: fixing the number of buckets, enforcing fairness constraints, and focusing on top-k rankings.
The framework applies to different aggregation settings, including the Optimal Bucket Order Problem (OBOP), which seeks a consensus weak order minimizing the distance to the input rankings. The OBOP is solved using efficient optimization techniques and can handle instances of practical size. Numerical experiments, together with a real-world case study on Spanish university rankings based on performance indicators, show that the method produces stable and interpretable aggregated rankings, especially when ties are relevant and standard methods are less informative.
Palabras clave: Rank Aggregation Weak Orders Fairness Optimal Bucket Order problem