Instance Selection Using Evolutionary Algorithms: An Experimental Study - Advanced Techniques in Knowledge Discovery and Data Mining - page 141

The objective of the EA is to maximize the fitness function defined, i.e.,
maximize the classification performance and minimize the number of instances
obtained. In the experiments, we have considered Į = 0.5.
5.6 Methodology for the Experiments
In this section, we present the methodology followed for the experiments. Section
5.6.1 describes the data sets used, Section 5.6.2 explains the partitions of the data
sets that were considered for applying the algorithms, and finally, Section 5.6.3
introduces the parameters associated with the algorithms.
5.6.1 Data Sets
A different group of data sets have been contemplated for each problem.
5.6.1.1 Instance Selection - Prototype Selection
We have evaluated 10 classical data sets used in machine learning for the PS [39]
shown in Table 5.1.
Cleveland : This database contains 76 attributes, but all published experiments
refer to using a subset of 13 of them. In particular, the Cleveland database is the
only one that has been used by machine learning researchers to this date. The
“goal” field refers to the presence of heart disease in the patient. It is integer-
valued from 0 (no presence) to 4. Experiments with the Cleveland database have
(forbreaking…)
Table 5.1. Data sets for IS-PS.
Data set Num. instances Num. features Num. classes
Cleveland 297 13 2
Glass 214 9 6
Iris 150 4 3
LED24Digit 200 24 10
LED7Digit 500 7 10
Lymphography 148 18 4
Monk 432 6 2
Pima 768 8 2
Wine 178 13 3
Wisconsin 683 9 2
concentrated on simply attempting to distinguish presence (values 1, 2, 3, 4) from
absence (value 0).