Fuzzy Jaccard Index: A robust comparison of ordered lists

We propose Fuzzy Jaccard Index (Fuji) — a scale-invariant score for similarity assessment of two ranked/ordered lists. Fuji improves upon the index by incorporating membership function that takes into account particular ranks, thus producing both more stable and accurate estimates. provide theoretical insights properties as well an efficient algorithm computing it. also present empirical evidence its performance in different synthetic scenarios. Finally, we demonstrate utility typical machine learning setting comparing feature ranking lists, relevant to given task. In many practical applications, originating from high-dimensional domains, where only small percentage whole space might be relevant, robust confident leads interpretable findings, computation good predictive performance. such cases, correctly distinguishes between existing approaches, while being than benchmark scores.