Computational Intelligence - February 2013 - 74

Table 3 Ensemble network performance without ten fold cross validation.
Classifiers

Hidden neurons

TP [%]

Tn [%]

aCCuraCy [%]

4
5

111, 39, 94, 31
111, 39, 94, 31, 33

47.0
45.0

39.0
43.0

86.0
88.0

6

111, 39, 94, 31, 33

46.0

33.0

79.0

10

111, 39, 94, 31, 33, 62, 87, 97, 149, 8

49.0

37.0

86.0

12

111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9, 10

45.0

39.0

84.0

13

111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9,
10, 45
111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9,
10, 45, 148, 7, 23, 130, 134

47.0

45.0

92.0

46.0

46.0

92.0

18
22

111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9,
10, 45, 148, 7, 23, 130, 134, 4, 25, 29, 118

48.0

33.0

81.0

25

111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9, 10,
45, 148, 7, 23, 130, 134, 4, 25, 29, 118, 16,
72, 137
111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9, 10,
45, 148, 7, 23, 130, 134, 4, 25, 29, 118, 16,
72, 137, 17, 61

49.0

32.0

81.0

47.0

44.0

91.0

27

30

111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9, 10,
45, 148, 7, 23, 130, 134, 4, 25, 29, 118, 16,
72, 137, 17, 61, 12, 63, 146

48.0

38.0

86.0

35

111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9,
10, 45, 148, 7, 23, 130, 134, 4, 25, 29, 118,
16, 72, 137, 17, 61, 12, 63, 146, 20, 21, 54,
67, 86

47.0

39.0

86.0

36

111, 39, 94, 31, 33, 62, 87, 97, 149, 8, 9,
10, 45, 148, 7, 23, 130, 134, 4, 25, 29, 118,
16, 72, 137, 17, 61, 12, 63, 146, 20, 21, 54,
67, 86, 6

47.0

43.0

90.0

39

111,39,94,31,33,62,87,97,149,8,9,10,45,148,
7,23,130,134,4,25,29,118,16,72,137,17,61,12,
63,146,20,21,54,67,86,6,36,24,27

47.0

43.0

90.0

as well [37]. A comparison of the techniques employed in this paper together
with the results for breast mass classification obtained by other researchers is
presented in Table 1.
Table 4 Accuracy for Adaboostm1
MLP classifier.
Hidden
neurons

TP [%]

Tn [%]

aCCuraCy
[%]

8

93.0

87.0

90.0

11

86.0

87.0

86.5

16

87.0

86.0

86.5

20
25

87.0

87.0

93.0

87.0

87.0
90.0

40

88.0

83.0

85.5

51

87.0

87.0

87.0

62
88

90.0

85.0

89.0

86.0

87.5
87.5

98

91.0

85.0

88.0

101

92.0

86.0

89.0

107
111
150

74

91.0
91.0
88.0

86.0
87.0
87.0

Fig. 7 shows the performance of the
proposed technique in comparison to
recently published techniques. It is difficult to compare results from one technique to another due to variances in
configuration parameters, research
methodology, datasets as well as other
metrics including time and computing
resources required to reach a classification. Fig. 7 however does contain
research using several techniques that
have been applied to exactly the same
dataset (ensemble MLP, MLP, Adaboost
M1 ensemble, clustered ensemble [35]
together with the proposed technique)

in order to ensure that an appropriate
performance evaluation can be undertaken of different classification techniques. The performance evaluation of
all techniques that has been tested on
the DDSM [21] has been included in
Fig. 7 in order to provide as fair a comparison as possible.
Examination of Fig. 7 indicates that
the performance of the proposed technique is comparable and better than that
achieved by a number of existing techniques. A direct comparison with our
earlier work on a clustered ensemble [35]
indicates that the proposed technique has
performed better on the mass dataset.
While performing the experiments
it was noted that some experiments
took much longer than others to complete. The Adaboost M1 experiments
ran for considerably longer than any
other technique.
As can be seen in Fig. 7, the proposed technique has improved the
classification accuracy, however to see
if the improvement was significant, an
ANOVA analysis was conducted. A 5%
confidence level was used. The
ANOVA summary and analysis are
presented in Tables 5 and 6.
As shown in Table 6, the p-value is significantly below the required 5% confidence level that confirms that the
improvement in performance of the proposed ensemble technique is statistically
significant. Table 5 (ANOVA summary)
shows that the variance for the ensemble
technique is much higher than that of the
MLP. This is because of the following reasons. The classifiers for the MLP were
chosen on the basis of being the highest
performers from the candidate classifier
pool. The classifiers for the ensemble
technique came from a much smaller candidate pool and the variation was therefore
higher as all candidates were chosen.

Table 5 ANOVA summary.
GrouPs

CounT

sum

averaGe

varianCe

MLP

38

3195.5

84.09

0.6197

ProPosed enseMbLe

38

3674.0

96.68

1.3706

89.0

enseMbLe MLP

38

3266.0

85.95

12.4296

87.5

AdAboost MLP

38

3337.5

87.83

0.9497

88.5

IEEE ComputatIonal IntEllIgEnCE magazInE | FEbruary 2013



Table of Contents for the Digital Edition of Computational Intelligence - February 2013

Computational Intelligence - February 2013 - Cover1
Computational Intelligence - February 2013 - Cover2
Computational Intelligence - February 2013 - 1
Computational Intelligence - February 2013 - 2
Computational Intelligence - February 2013 - 3
Computational Intelligence - February 2013 - 4
Computational Intelligence - February 2013 - 5
Computational Intelligence - February 2013 - 6
Computational Intelligence - February 2013 - 7
Computational Intelligence - February 2013 - 8
Computational Intelligence - February 2013 - 9
Computational Intelligence - February 2013 - 10
Computational Intelligence - February 2013 - 11
Computational Intelligence - February 2013 - 12
Computational Intelligence - February 2013 - 13
Computational Intelligence - February 2013 - 14
Computational Intelligence - February 2013 - 15
Computational Intelligence - February 2013 - 16
Computational Intelligence - February 2013 - 17
Computational Intelligence - February 2013 - 18
Computational Intelligence - February 2013 - 19
Computational Intelligence - February 2013 - 20
Computational Intelligence - February 2013 - 21
Computational Intelligence - February 2013 - 22
Computational Intelligence - February 2013 - 23
Computational Intelligence - February 2013 - 24
Computational Intelligence - February 2013 - 25
Computational Intelligence - February 2013 - 26
Computational Intelligence - February 2013 - 27
Computational Intelligence - February 2013 - 28
Computational Intelligence - February 2013 - 29
Computational Intelligence - February 2013 - 30
Computational Intelligence - February 2013 - 31
Computational Intelligence - February 2013 - 32
Computational Intelligence - February 2013 - 33
Computational Intelligence - February 2013 - 34
Computational Intelligence - February 2013 - 35
Computational Intelligence - February 2013 - 36
Computational Intelligence - February 2013 - 37
Computational Intelligence - February 2013 - 38
Computational Intelligence - February 2013 - 39
Computational Intelligence - February 2013 - 40
Computational Intelligence - February 2013 - 41
Computational Intelligence - February 2013 - 42
Computational Intelligence - February 2013 - 43
Computational Intelligence - February 2013 - 44
Computational Intelligence - February 2013 - 45
Computational Intelligence - February 2013 - 46
Computational Intelligence - February 2013 - 47
Computational Intelligence - February 2013 - 48
Computational Intelligence - February 2013 - 49
Computational Intelligence - February 2013 - 50
Computational Intelligence - February 2013 - 51
Computational Intelligence - February 2013 - 52
Computational Intelligence - February 2013 - 53
Computational Intelligence - February 2013 - 54
Computational Intelligence - February 2013 - 55
Computational Intelligence - February 2013 - 56
Computational Intelligence - February 2013 - 57
Computational Intelligence - February 2013 - 58
Computational Intelligence - February 2013 - 59
Computational Intelligence - February 2013 - 60
Computational Intelligence - February 2013 - 61
Computational Intelligence - February 2013 - 62
Computational Intelligence - February 2013 - 63
Computational Intelligence - February 2013 - 64
Computational Intelligence - February 2013 - 65
Computational Intelligence - February 2013 - 66
Computational Intelligence - February 2013 - 67
Computational Intelligence - February 2013 - 68
Computational Intelligence - February 2013 - 69
Computational Intelligence - February 2013 - 70
Computational Intelligence - February 2013 - 71
Computational Intelligence - February 2013 - 72
Computational Intelligence - February 2013 - 73
Computational Intelligence - February 2013 - 74
Computational Intelligence - February 2013 - 75
Computational Intelligence - February 2013 - 76
Computational Intelligence - February 2013 - 77
Computational Intelligence - February 2013 - 78
Computational Intelligence - February 2013 - 79
Computational Intelligence - February 2013 - 80
Computational Intelligence - February 2013 - Cover3
Computational Intelligence - February 2013 - Cover4
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202311
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202308
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202305
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202302
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202211
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202208
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202205
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202202
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202111
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202108
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202105
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202102
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202011
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202008
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202005
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_202002
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201911
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201908
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201905
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201902
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201811
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201808
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201805
https://www.nxtbook.com/nxtbooks/ieee/computationalintelligence_201802
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_winter17
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_fall17
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_summer17
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_spring17
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_winter16
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_fall16
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_summer16
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_spring16
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_winter15
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_fall15
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_summer15
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_spring15
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_winter14
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_fall14
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_summer14
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_spring14
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_winter13
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_fall13
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_summer13
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_spring13
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_winter12
https://www.nxtbook.com/nxtbooks/ieee/computational_intelligence_fall12
https://www.nxtbookmedia.com