Graphics Reference
In-Depth Information
Table 9.8
Wilcoxon test results in accuracy
C4.5 Data Squeezer KNN Naıve Bayes PUBLIC Ripper
+±+
± +±+±+±+±
1R
1123
23
2191
9 1 2111
Ameva
14 29 17
29
8269
29
13 29
9 29
Bayesian
195
26
212210 1217
CACC
92816
29
218589
29 426
CADD
011
22 010
1 0600
CAIM
16 29 16
29
11 28 10
29
16 29
11 28
Chi2
13 29 4
26
6279
29
11 29
19 29
ChiMerge
17 29 18
29
13 28 10
29
17 29 928
ClusterAnalysis 1100
12
524671 1220
DIBD
6218
29 292
8 9 315
Distance
13 29 16
29
2177613 28 213
EqualFrequency 10 27 3
21
18 29 9
29
10 26
11 27
EqualWidth
7202
18
11 28 8
29 6 0927
Extended Chi2 9274
26
319376 5225
FFD
5150
5
20 28 8
29 1 310 27
FUSINTER
21 29 9
29
12 28 15
29
20 29
11 29
HDD
1180
14
423580 4726
HellingerBD
10 27 4
22
7267810 26 626
Heter-Disc
099
29 020
3 0 1110
ID3
1100
5
522480 1526
IDD
1103
23
421240 2116
Khiops
12 27 3
18
18 29 9
29 9 711 29
MDLP
14 29 14
29
3228
29
15 29 216
Modified Chi2
11 27 3
21
17 28 10
29
9
29
23 29
MODL
12 28 5
23
14 28 9
29
10 28
17 29
MVD
1155
29 181
7 0 9113
PKID
5150
6
27 29 9
29 1 315 29
UCPD
14 29 7
26
4172514 28 319
USD
1133
19
6236
29 1 9725
Zeta
14 29 17
29
4209
29
14 29 727
comparisons involved in the Wilcoxon test among all discretizers and measures, for
number of intervals and inconsistency rate, accuracy and kappa respectively. Again,
the individual comparisons between all possible discretizers are exhibited in the
aforementioned URL, where a detailed report of statistical results can be found for
each measure and classifier. Tables
9.7
,
9.8
and
9.9
summarize, for each method in the
rows, the number of discretizers outperformed by using the Wilcoxon test under the
