Supervised Machine Learning Based Noninvasive Prediction of Atrial Flutter Mechanism from P-to-P Interval Variability under Imbalanced Dataset Conditions
Table 8
Median and standard deviation values of both filter and feature selection methods on all synthetic techniques at 400% data augmentation.
Median of p value by filter method
Median of feature scores by wrapper method
Features
SMOTE
Modified-SMOTE
Smoothed-bootstrap
SMOTE
Modified-SMOTE
Smoothed-bootstrap
LDA
LOG
SVM
LDA
LOG
SVM
LDA
LOG
SVM
Mean
0.54
0.55
0.55
0.7
0.7
0.8
0.6
0.7
0.7
0.5
0.6
0.8
Median
0.88
0.91
0.84
0.7
0.7
0.8
0.5
0.7
0.7
0.6
0.6
0.7
Mode
0.95
0.94
0.87
0.5
0.7
0.7
0.4
0.7
0.7
0.5
0.7
0.6
Std
0.86
0.40
0.34
0.7
0.6
0.8
0.7
0.7
0.8
0.7
0.7
0.9
Var
0.86
0.40
0.34
0.7
0.6
0.8
0.8
0.7
0.8
0.7
0.65
0.8
Skew
0.11
0.11
0.28
0.6
0.7
0.8
0.6
0.7
0.7
0.6
0.7
0.7
Kurt
0.51
0.74
0.76
0.7
0.8
0.9
0.7
0.3
0.8
0.7
0.6
0.7
PDmin
0.22
0.15
0.13
0.7
0.7
0.9
0.8
0.9
0.9
0.8
0.9
1
PDmax
0.27
0.58
0.46
0.5
0.6
0.7
0.6
0.6
0.7
0.5
0.6
0.7
Interval
0.01
0.01
0.01
1
1
1
1
1
1
0.9
1
1
Standard deviation of p value by filter method
Standard deviation of feature scores by wrapper method
Mean
0.055
0.068
0.109
0.114
0.132
0.082
0.113
0.095
0.116
0.115
0.155
0.071
Median
0.051
0.071
0.095
0.110
0.088
0.089
0.107
0.096
0.090
0.090
0.111
0.096
Mode
0.049
0.068
0.091
0.137
0.090
0.110
0.117
0.123
0.138
0.114
0.130
0.188
Std
0.078
0.085
0.146
0.089
0.122
0.116
0.122
0.130
0.116
0.106
0.083
0.099
Var
0.078
0.085
0.146
0.077
0.149
0.084
0.064
0.133
0.085
0.083
0.116
0.072
Skew
0.071
0.094
0.148
0.125
0.124
0.177
0.100
0.082
0.160
0.130
0.097
0.145
Kurt
0.188
0.180
0.188
0.123
0.176
0.102
0.122
0.141
0.089
0.127
0.112
0.118
PDmin
0.061
0.038
0.056
0.112
0.189
0.169
0.115
0.034
0.145
0.139
0.078
0.074
PDmax
0.053
0.089
0.074
0.167
0.098
0.095
0.105
0.115
0.093
0.146
0.126
0.134
Interval
0.001
0.00
0.002
0.010
0.031
0.010
0.026
0.038
0.024
0.050
0.014
0.080
The relevant features have been emphasized, and the use of bold font indicates that certain sets of features have achieved greater significance than the arbitrary threshold of 0.8.
