Affiliations: University of Bucharest, Faculty of Mathematics and
Computer Science, Academiei 14, 010014, Bucharest, Romania. E-mail:
ldinu@funinf.cs.unibuc.ro | University of Bucharest, Faculty of Mathematics and
Computer Science, Academiei 14, 010014, Bucharest, Romania. E-mail:
mpopescu@phobos.cs.unibuc.ro
Note: [] Address for correspondence: University of Bucharest, Faculty of
Mathematics and Computer Science, Academiei 14, 010014, Bucharest, Romania
Abstract: The multi-criteria decision making process can be summarized as
follows. Given a pattern d and a set C = {c_1,
c_2, …, c_m} of allmpossible
categories of d, we are interested in predicting its class by using a set of n
classifiers l_1, l_2, …,
l_n. Each classifier produces a ranking of categories. In
this paper we propose and test a decision method which combines the rankings by
using a particular method, called rank distance categorization. This method is
actually based on the rank distance, a metric which was successfully used in
computational linguistics and bioinformatics. We define the method, present
some of its mathematical and computational properties and we test it on the
digit dataset consisting of handwritten numerals ('0', …, '9') extracted
from a collection of Dutch utility maps. We compare our experimental results
with other reported experiments which used the same dataset but different
combining methods.
