Round Randomized Learning Vector Quantization for Brain Tumor Imaging
Table 2
Accuracy rates for LVQ classifiers before and after applying the proposed work and the results of round-off and multirandomization methods individually.
Classifier
Datasets
Brain images data (PPUKM)
Segmented challenge (UCI)
Segment test (UCI)
Image segmentation (UCI)
Average
STD
LVQ1
BRR
82.24
86.22
84.36
89.61
85.61
3.12
Only random
87
87
85
89
86
1.63
Only round
76
77
77
78
77
0.82
ARR
88.16
91.11
91.36
91.34
90.49
1.56
Multi pass LVQ
BRR
78.29
87.78
89.3
90.19
86.39
5.49
Only random
83
85.5
81
87.7
85.4
2.9
Only round
70
64
68
69
71
2.63
ARR
84.87
93.33
93
92.21
90.85
4.02
HLVQ
BRR
80.92
92.89
91.36
91.92
89.27
5.6
Only random
83
92
87
92
89
4.36
Only round
74
74
70
75
73
2.22
ARR
84.21
96
94.24
95.24
92.42
5.52
OLVQ3
BRR
75.66
86.89
88.07
89.61
85.06
6.36
Only random
85.75
90.45
91.1
89.1
89
2.38
Only round
76
73.9
62.5
63.5
69
6.96
ARR
83.55
90.44
90.95
92.21
89.29
3.9
ALL LVQ siblings
Ave. BRR
79.2775
88.445
88.2725
90.3325
86.58
4.96
Ave. only random
84.6875
88.7375
86.025
89.45
87.35
2.8175
Ave. only round
74
72
69.4
71.4
71.7
1.92
Ave. ARR
85.1975
92.72
92.3875
92.75
90.76
3.71
(i) BRR means before rounding the distance function results and multirandomizing data sets. (ii) ARR means after rounding the distance function results and multirandomizing data sets.
