Graphics Reference
In-Depth Information
Table 8.1
(continued)
Complete name
Abbr. name
Reference
Instance selection based on classification
contribution
ISCC
[
17
]
Bayesian instance selection
EVA
[
56
]
Reward-Punishment editing
RP-Edit
[
57
]
Complete cross validation functional pro-
totype selection
CCV
[
87
]
Sequential reduction algorithm
SeqRA
[
132
]
Local support vector machines noise reduc-
tion
LSVM
[
145
]
No name specified
Bien
[
13
]
Reverse nearest neighbor reduction
RNNR
[
35
]
Border-Edge pattern selection
BEPS
[
105
]
Class boundary preserving algorithm
CBP
[
122
]
Cluster-Based instance selectio
CBIS
[
34
]
RDCL profiling
RDCL
[
38
]
Multi-Selection genetic algorithm
MSGA
[
73
]
Ant colony prototype reduction
Ant-PR
[
118
]
Spectral instance reduction
SIR
[
123
]
Competence
enhancement
by
Ranking-
CRIS
[
37
]
based instance selection
Discriminative prototype selection
D-PS
[
14
]
Adaptive threshold-based instance selec-
tion algorithm
AT I S A
[
24
]
InstanceRank based on borders for instance
selection
IRB
[
85
]
Visualization-Induced self-organizing map
for prototype reduction
VISOM
[
106
]
Support vector oriented instance selection
SVOIS
[
154
]
Dominant set clustering prototype selec-
tion
DSC
[
157
]
Fuzzy rough prototype selection
FRPS
[
159
]
Figure
8.3
illustrates the categorization following a hierarchy based on this order:
type of selection, direction of search and evaluation of the search. It allows us to
distinguish among families of methods and to estimate the size of each one.
One of the objectives in this chapter is to highlight the best methods depending on
their properties, taking into account that we are conscious that the properties could
determine the suitability of use of a specific scheme. To do this, in Sect.
8.6
, we will
conclude which methods perform best for each family considering several metrics
of performance.
