Graphics Reference
In-Depth Information
Table 10.6
Average results and standard deviations of test accuracy obtained
Data set Ant Miner
CORE
HIDER
SGERD
TARGET
Mean
SD
Mean
SD
Mean
SD
Mean
SD
Mean
SD
HAB
72.55
5.27
72.87
4.16
75.15
4.45
74.16
2.48
71.50
2.52
IRI
96.00
3.27
92.67
4.67
96.67
3.33
96.67
3.33
92.93
4.33
BAL
70.24
6.21
70.08
7.11
69.60
3.77
75.19
6.27
75.62
7.27
NTH
90.76
6.85
90.76
5.00
90.28
7.30
88.44
6.83
86.79
5.83
MAM
81.48
7.38
77.33
3.55
82.30
6.50
74.11
5.11
79.65
2.11
BUP
57.25
7.71
61.97
4.77
65.83
10.04
57.89
3.41
65.97
1.41
MON
97.27
2.65
88.32
8.60
97.27
2.65
80.65
4.15
96.79
5.15
CAR
77.26
2.59
79.40
3.04
70.02
0.16
67.19
0.70
77.71
2.70
ECO
58.58
9.13
64.58
4.28
75.88
6.33
72.08
7.29
65.49
4.29
LED
55.32
4.13
27.40
4.00
68.20
3.28
40.00
6.75
32.64
6.75
PIM
66.28
4.26
73.06
6.03
73.18
6.19
73.71
3.61
73.02
6.61
GLA
53.74
12.92
45.74
9.36
64.35
12.20
48.33
5.37
44.11
5.37
WIS
90.41
2.56
92.38
2.31
96.05
2.76
92.71
3.82
95.75
0.82
TAE
64.61
5.63
70.35
3.77
69.93
4.73
69.93
4.73
69.50
2.73
WIN
92.06
6.37
94.87
4.79
82.61
6.25
87.09
6.57
82.24
7.57
CLE
57.45
5.19
53.59
7.06
55.86
5.52
44.15
4.84
52.99
1.84
HOU
93.56
3.69
97.02
3.59
97.02
3.59
97.02
3.59
96.99
0.59
LYM
73.06
10.98
65.07
15.38
72.45
10.70
72.96
13.59
75.17
10.59
VEH
53.07
4.60
36.41
3.37
63.12
4.48
51.19
4.85
49.81
5.85
BAN
59.18
6.58
64.23
4.23
62.15
8.51
62.71
4.17
67.32
6.17
GER
66.90
3.96
69.30
1.55
70.40
4.29
66.70
1.49
70.00
0.49
AUT
53.74
7.79
32.91
6.10
62.59
13.84
50.67
10.27
42.82
13.27
DER
81.16
7.78
31.03
1.78
87.45
3.26
69.52
4.25
66.15
4.25
SON
71.28
5.67
53.38
1.62
52.90
2.37
73.45
7.34
74.56
8.34
Average 72.22
5.97
66.86
5.01
75.05
5.69
70.27
5.20
71.06
4.87
description), by including a node called Stat-Clas-Friedman in the KEEL experiment.
Here, we include the information provided by this statistical module:
Table
10.7
shows the obtained average rankings across all data sets following the
Friedman procedure for each method. They will be useful to calculate the
p
-value
and to detect significant differences between the two methods.
Table
10.8
depicts the results obtained from the use of the Friedman and Iman-
Davenport test. Both, the statistics and
p
-values are shown. As we can see, a level
of significance
α
=
.
10 is needed in order to consider that differences among the
methods exist. Note also that the
p
-value obtained by the Iman-Davenport test is
lower than that obtained by Friedman, this is always true.
0
