A Feature Selection Algorithm Integrating Maximum Classification Information and Minimum Interaction Feature Dependency Information
Table 5
Average classification accuracy (%) of SVM classifier.
Data set
NDCRFS
MIM
IG-RFE
IWFS
CMIM
DWFS
CIFE
Lymphography
45.147
42.499
43.329
41.45
42.825
43.329
42.825
Dermatology
98.317
93.777
93.824
93.283
94.079
97.761
93.53
Cardiotocography
98.448
98.401
98.401
98.401
98.401
98.401
98.401
Pendigits
63.331
63.331
63.331
55.35
59.741
56.979
57.219
Lung
84.788
77.89
78.391
77.891
86.203
85.311
77.402
Carcinom
87.964
50.998
25.028
50.447
51.545
55.773
20.915
Nci9
76.512
78.119
76.69
62.595
74.429
57.929
58.821
PCMAC
85.589
85.588
85.486
82.194
85.333
85.382
80.394
Pixraw10P
92.0
91.0
91.0
91.0
91.0
91.0
91.0
SMK-CAN-187
70.982
70.569
62.532
71.593
65.32
71.053
57.255
Lymphoma
85.5
81.278
79.611
67.056
81.972
72.194
86.194
COIL20
68.352
63.886
62.067
52.824
55.933
48.638
40.905
Average accuracy rate
79.213
73.363
71.641
70.226
73.898
71.979
65.333
Wins/Ties/Losses
10/1/1
12/0/0
12/0/0
11/0/1
10/0/2
11/0/1
The “Average” column gives the average accuracy value of the feature selection algorithm over all datasets. Bold represents the highest average classification prediction under this dataset.