Michael Rademaker, Ph.D. student at KERMIT, Faculty of Bioscience
Engineering, Ghent University.

Preference for oral presentation.

Conflicting evaluations in Multiple Criteria Decision Making:
quantification and remediation of non-monotonicity
Michael Rademaker, Bernard De Baets and Hans De Meyer.

Abstract: Multi-criteria data sets for the supervised training of
aggregation operators are prone to a special kind of noise or inaccuracy
in the form of non-monotonicity. This non-monotonicity restricts the
extent to which an aggregation operator can be fitted through the data.
Three quantifications of the monotonicity of a data set are discussed, as
well as two heuristics to clean up a data set while greedily maximising
the monotonicity of the data set as quantified by two of the three
measures. Additionally, the relation to the maximum independent set
problem shown.

Keywords: aggregation, multi-criteria data, maximum independent set, data
cleansing, monotone approximation