Serviços Personalizados
Journal
Artigo
Indicadores
- Citado por SciELO
- Acessos
Links relacionados
- Citado por Google
- Similares em SciELO
- Similares em Google
Compartilhar
TecnoLógicas
versão impressa ISSN 0123-7799versão On-line ISSN 2256-5337
Resumo
REYNA-HERNANDEZ, Randy e ROSETE, Alejandro. Extension of the Concept of Utopia for Rank Aggregation Problem Without Ties. TecnoL. [online]. 2021, vol.24, n.51, pp.16-31. Epub 04-Out-2021. ISSN 0123-7799. https://doi.org/10.22430/22565337.1788.
The use of rankings and how to aggregate or summarize them has received increasing attention in various fields: bibliometrics, web search, data mining, statistics, educational quality, and computational biology. For the Optimal Bucket Order Problem, the concept of Utopian Matrix was recently introduced: an ideal and not necessarily feasible solution with an unsurpassed quality for the feasible solutions of the problem. This work proposes an extension of the notion of Utopian Matrix to the Rank Aggregation Problem in which ties are not allowed between elements in the output ranking. Beyond the extension that is direct, the work focuses on studying its usefulness as an idealization or super optimal solution. As the Rank Aggregation Problem can be solved exactly based on its definition as an Integer Linear Programming Problem, an experimental study is presented where it is analyzed the relationship that exists between utopian (and anti utopian) values and the optimal solution in several instances solved by using the open source software SCIP. Among the 47 instances analyzed, in 19 the Utopian Value turned out to be equal to the optimal value (40.43 % feasibility) and in 18 the Anti Utopian Value also turned out to be feasible (38.00 %). This experimental study demonstrates the usefulness of utopian and anti utopian values to be considered as extreme values in the Rank Aggregation Problem, thus being able to find higher and lower bounds for optimization very quickly.
Palavras-chave : Rank aggregation; integer linear programming; data mining; approximation algorithms.