We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.AI

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Artificial Intelligence

Title: Learning MR-Sort Models from Non-Monotone Data

Abstract: The Majority Rule Sorting (MR-Sort) method assigns alternatives evaluated on multiple criteria to one of the predefined ordered categories. The Inverse MR-Sort problem (Inv-MR-Sort) computes MR-Sort parameters that match a dataset. Existing learning algorithms for Inv-MR-Sort consider monotone preferences on criteria. We extend this problem to the case where the preferences on criteria are not necessarily monotone, but possibly single-peaked (or single-valley). We propose a mixed-integer programming based algorithm that learns the preferences on criteria together with the other MR-Sort parameters from the training data. We investigate the performance of the algorithm using numerical experiments and we illustrate its use on a real-world case study.
Subjects: Artificial Intelligence (cs.AI); Other Computer Science (cs.OH)
Cite as: arXiv:2107.09668 [cs.AI]
  (or arXiv:2107.09668v1 [cs.AI] for this version)

Submission history

From: Pegdwende Minoungou [view email]
[v1] Tue, 20 Jul 2021 13:51:16 GMT (242kb,D)

Link back to: arXiv, form interface, contact.