A Feature Selection Algorithm Integrating Maximum Classification Information and Minimum Interaction Feature Dependency Information
Table 3
Average classification accuracy (%) of KNN classifier.
Data set
NDCRFS
MIM
IG-RFE
IWFS
CMIM
DWFS
CIFE
Lymphography
38.3
34.78
35.59
35.59
34.88
35.28
34.78
Dermatology
97.769
92.164
92.164
88.512
90.79
96.68
87.139
Cardiotocography
98.589
98.401
98.401
98.401
98.401
98.589
98.401
Pendigits
97.919
97.145
97.145
97.238
97.505
98.159
97.625
Lung
88.636
88.064
83.712
76.391
81.678
87.681
74.922
Carcinom
85.48
68.037
32.255
60.035
65.84
67.026
31.952
Nci9
76.69
75.44
74.012
69.024
76.119
48.429
57.25
PCMAC
87.648
85.538
86.155
82.348
84.765
85.743
78.952
Pixraw10P
93.0
88.0
91.0
88.0
92.0
88.0
92.0
SMK-CAN-187
70.014
68.393
69.004
70.0
65.747
68.421
58.876
Lymphoma
95.667
84.722
84.75
69.806
90.083
72.056
82.833
COIL20
84.662
80.733
79.743
71.667
77.114
72.024
60.652
Average accuracy rate
88.734
84.24
76.994
75.584
83.64
76.507
71.28
Wins/Ties/Losses
12/0/0
12/0/0
12/0/0
12/0/0
12/0/0
12/0/0
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.