Information Technology Reference
In-Depth Information
Tabl e 3 . Results of Tukey's HSD post-hoc test for the “Vote” dataset
Dependent
Variable
(I)
Approach
(J)
Approach
Mean Difference
(I-J)
Std.
Error
95% Confidence Interval
Sig.
Lower Bound
Upper Bound
G
16 4 3 3
G
16 3 3 3
-.08621
.1105
.863
-.3766
.2042
G
16 1 3 3
-.22414
.1105
.187
-.5145
.0662
Fitness
training
.08621
.1105
.863
-.2042
.3766
G
16 3 3 3
G
16 4 3 3
G
16 1 3 3
-.13793
.1105
.599
-.4283
.1524
.22414
.1105
.187
-.0662
.5145
G
16 1 3 3
G
16 4 3 3
G
16 3 3 3
.13793
.1105
.599
-.1524
.4283
G
16 4 3 3
G
16 3 3 3
.20690
.1879
.690
-.2866
.7004
G
16 1 3 3
.41379
.1879
.132
-.0797
.9073
G
16 3 3 3
G
16 4 3 3
Fitness
testing
-.20690
.1879
.690
-.7004
.2866
G
16 1 3 3
.20690
.1879
.690
-.2866
.7004
G
16 1 3 3
G
16 4 3 3
-.41379
.1879
.132
-.9073
.0797
G
16 3 3 3
-.20690
.1879
.690
-.7004
.2866
Tabl e 4 . Descriptive statistical results for 20 executions of EvoBANE with grammars
G
8444
,
G
8344
,
G
8144
and a genetic algorithm for the “Postoperative” dataset
Std.
Deviation
Std.
Error
95% Confidence Interval for Mean
N
Mean
Minimum Maximum
Lower Bound
Upper Bound
G
8444
20
80.00
.000
.000
80.00
80.00
80
80
G
8344
20
81.50
4.617
1.032
79.34
83.66
80
95
Fitness
training
G
8144
20
80.00
.000
.000
80.00
80.00
80
80
GA
20
75.00
.000
.000
75.00
75.00
75
75
Total
80
79.13
3.354
.375
78.38
79.87
75
95
G
8444
20 72.6667
6.3614
1.4225
69.6894
75.6439
63.3333
80
G
8344
20 73.6667
7.0835
1.5839
70.3515
76.9818
63.3333
80
Fitness
testing
G
8144
20 73.5000
7.1308
1.5945
70.1627
76.8373
63.3333
80
GA
20
60
.0000
.0000
60
60
60
60
Total
80 69.9583
8.2249
.9196
68.1280
71.7887
60
80
Tabl e 5 . Results of the ANOVA test for the “Postoperative” dataset
Sum of Squares
df
Mean Square
F
Sig.
Between
30.000
2
15.000
2.111
.130
Fitness
training
Within Groups
405.000
57
7.105
Total
435.000
59
Between
11.481
2
5.741
.122
.886
Fitness
testing
Within Groups
2688.333
57
47.164
Total
2699.815
59
Therefore the reduction of the search space by the grammars
G
8344
and
G
8144
does
not affect the performance of the BNs generated by EvoBANE.
4 Conclusions
The EvoBANE system has been presented as an evolutionary approach that
automatically generates Bayesian networks for solving classification problems.
Search WWH ::
Custom Search