Graphics Reference
In-Depth Information
Prototype
Selection
Algorithm
Training
Data Set
(TR)
Instances
Selected
(S)
Instance-based
Classifier
Test
Data Set
(TS)
Fig. 8.1
PS process
Training Set
Selection
Algorithm
Training
Data Set
(TR)
Instances
Selected
(S)
Data Mining
Algorithm
Model
Obtained
Test
Data Set
(TS)
Fig. 8.2
TSS process
TSS methods are defined in a similar way. They are known as the application
of IS methods over the training set used to build any predictive model. Thus, TSS
can be employed as a way to improve the behavior of predictive models, precision
and interpretability [
135
]. Figure
8.2
shows the basic steps of processing a decision
tree (C4.5) on the TSS. Among others, ANNs [
51
,
94
,
160
], SVMs [
31
], decision
trees [
21
,
85
]; and even in other learning paradigms such as regression [
8
,
154
],
time series forecasting [
79
,
170
], subgroup discovery [
22
,
23
], imbalanced learning
[
11
,
65
,
66
,
75
,
110
], multiple-instance learning [
30
,
58
], and semi-supervised learn-
ing [
80
].
