WIRE: A Weighted Item Removal Method for Unsupervised Rank Aggregation
Rank aggregation deals with the problem of fusing multiple ranked lists of elements into a single aggregate list with improved element ordering. Such cases are frequently encountered in numerous applications across a variety of areas, including bioinformatics, machine learning, statistics, informati...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-06-01
|
| Series: | Algorithms |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1999-4893/18/6/362 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Rank aggregation deals with the problem of fusing multiple ranked lists of elements into a single aggregate list with improved element ordering. Such cases are frequently encountered in numerous applications across a variety of areas, including bioinformatics, machine learning, statistics, information retrieval, and so on. The weighted rank aggregation methods consider a more advanced version of the problem by assuming that the input lists are not of equal importance. In this context, they first apply ad hoc techniques to assign weights to the input lists, and then, they study how to integrate these weights into the scores of the individual list elements. In this paper, we adopt the idea of exploiting the list weights not only during the computation of the element scores, but also to determine which elements will be included in the consensus aggregate list. More specifically, we introduce and analyze a novel refinement mechanism, called WIRE, that effectively removes the weakest elements from the less important input lists, thus improving the quality of the output ranking. We experimentally demonstrate the effectiveness of our method in multiple datasets by comparing it with a collection of state-of-the-art weighted and non-weighted techniques. |
|---|---|
| ISSN: | 1999-4893 |