Graphics Reference
In-Depth Information
Table 5.9
Test accuracy, RLA results and p-values on data sets with attribute noise. Cases where
the baseline classifiers obtain more ranks than the OVO version in the Wilcoxon's test are indicated
with a star (*)
Uniform random attribute noise Gaussian attribute noise
C4.5 Ripper 5-NN C4.5 Ripper 5-NN
Base OVO Base OVO Base OVO Base OVO Base OVO Base OVO
0% 81.66 82.7 77.92 82.15 82.1 83.45 81.66 82.7 77.92 82.15 82.1 83.45
10% 80.31 81.65 76.08 80.85 79.81 81.34 80.93 81.67 76.53 81.12 80.91 82.52
20% 78.71 80.27 73.95 79.15 77.63 79.38 79.77 81.11 75.35 80.06 80.16 81.74
30% 76.01 78.25 71.25 77.06 74.68 76.46 79.03 80.4 74.46 78.93 78.84 80.77
40% 73.58 76.19 68.66 74.56 71.29 73.65 77.36 79.51 72.94 78.1 77.53 79.11
50% 70.49 73.51 65.5 71.66 67.72 70.07 75.29 78.03 71.57 76.27 76.02 77.72
0%
0.007
0.0002
0.093
0.007
0.0002
0.093
10%
0.0169
0.0003
0.091
0.1262
0.0004
0.0064
20%
0.0057
0.0003
0.0015
0.0048
0.0002
0.0036
30%
0.0043
0.0001
0.0112
0.0051
0.0003
0.0025
40%
0.0032
0.0001
0.0006
0.0019
0.0003
0.1262
50%
0.0036
0.0007
0.0011
0.0004
0.0008
0.0251
0% - - - - - -
10% 1.82 1.32 2.56 1.62 3.03 2.62 0.92 1.27 1.9 1.26 1.68 1.13
20% 3.88 3.11 5.46 3.77 5.72 5.03 2.4 1.99 3.49 2.59 2.51 2.09
30% 7.54 5.77 9.2 6.42 9.57 8.76 3.42 2.91 4.66 3.95 4.34 3.32
40% 10.64 8.25 12.81 9.64 13.73 12.08 5.67 4.03 6.74 4.96 6.03 5.38
50% 14.74 11.74 17.21 13.33 18.14 16.55 8.37 5.87
8.5
7.35 7.82 7.16
0%
-
-
-
-
-
-
10%
0.4781
0.5755
1.0000(*)
0.0766(*)
0.8519(*)
0.4115
20%
0.2471
0.1454
0.1354
0.8405
0.9108(*)
0.3905
30%
0.0304
0.0438
0.1672
0.6542
0.2627(*)
0.2627
40%
0.0569
0.0036
0.0111
0.1169
0.3905
0.9405
50%
0.0152
0.0064
0.0228
0.009
0.6542(*)
0.218
noise affects the attributes in a random and uniformway. This behavior is particularly
notable with the highest noise levels, where the effects of noise are expected to be
more detrimental.
