單獨BAI_T1, BDI_T1
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
NaiveBayes 0.914 0.914 0.667 0.750 0.600 0.870 0.833 0.909 0.545 0.812 1.600 4.800
LogisticRegression 0.908 0.908 0.718 0.737 0.700 0.876 0.867 0.886 0.595 0.828 1.900 4.500
SVM 0.878 0.878 0.696 0.615 0.800 0.829 0.895 0.773 0.541 0.781 2.600 3.800
MLP 0.878 0.878 0.632 0.667 0.600 0.844 0.826 0.864 0.478 0.781 1.800 4.600
RandomForest 0.863 0.863 0.611 0.688 0.550 0.848 0.812 0.886 0.467 0.781 1.600 4.800
KNN 0.863 0.863 0.571 0.667 0.500 0.839 0.796 0.886 0.423 0.766 1.500 4.900
XGBoost 0.849 0.849 0.615 0.632 0.600 0.831 0.822 0.841 0.447 0.766 1.900 4.500
DecisionTree 0.695 0.695 0.579 0.611 0.550 0.822 0.804 0.841 0.403 0.750 1.800 4.600
EEG_PWR_REL_BETA1_BRAIN_AVG （Beta1 全腦平均）
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EEG_PWR_REL_BETA1_BRAIN_AVG']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.935 0.935 0.750 0.750 0.750 0.886 0.886 0.886 0.636 0.844 2.000 4.400
NaiveBayes 0.920 0.920 0.667 0.750 0.600 0.870 0.833 0.909 0.545 0.812 1.600 4.800
LogisticRegression 0.909 0.909 0.700 0.700 0.700 0.864 0.864 0.864 0.564 0.812 2.000 4.400
SVM 0.902 0.902 0.851 0.741 1.000 0.914 1.000 0.841 0.789 0.891 2.700 3.700
RandomForest 0.887 0.887 0.629 0.733 0.550 0.860 0.816 0.909 0.502 0.797 1.500 4.900
KNN 0.862 0.862 0.800 0.800 0.800 0.909 0.909 0.909 0.709 0.875 2.000 4.400
XGBoost 0.847 0.847 0.647 0.786 0.550 0.872 0.820 0.932 0.540 0.812 1.400 5.000
DecisionTree 0.720 0.720 0.615 0.632 0.600 0.831 0.822 0.841 0.447 0.766 1.900 4.500
EEG_PWR_REL_BETA_T3T4_AVG （Beta_T3,T4顳葉平均）
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EEG_PWR_REL_BETA_T3T4_AVG']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.928 0.928 0.780 0.762 0.800 0.897 0.907 0.886 0.678 0.859 2.100 4.300
LogisticRegression 0.917 0.917 0.718 0.737 0.700 0.876 0.867 0.886 0.595 0.828 1.900 4.500
NaiveBayes 0.911 0.911 0.649 0.706 0.600 0.857 0.830 0.886 0.510 0.797 1.700 4.700
KNN 0.910 0.910 0.684 0.722 0.650 0.867 0.848 0.886 0.553 0.812 1.800 4.600
RandomForest 0.907 0.907 0.611 0.688 0.550 0.848 0.812 0.886 0.467 0.781 1.600 4.800
SVM 0.900 0.900 0.780 0.762 0.800 0.897 0.907 0.886 0.678 0.859 2.100 4.300
XGBoost 0.890 0.890 0.737 0.778 0.700 0.889 0.870 0.909 0.628 0.844 1.800 4.600
DecisionTree 0.793 0.793 0.718 0.737 0.700 0.876 0.867 0.886 0.595 0.828 1.900 4.500
‘EEG_PWR_REL_DELTA_FP1’, ‘EEG_PWR_REL_DELTA_FP2’ （雙側前額葉相對功率）
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=4
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EEG_PWR_REL_DELTA_FP1', 'EEG_PWR_REL_DELTA_FP2']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
RandomForest 0.922 0.922 0.684 0.722 0.650 0.867 0.848 0.886 0.553 0.812 1.800 4.600
SVM 0.907 0.907 0.756 0.680 0.850 0.867 0.923 0.818 0.635 0.828 2.500 3.900
MLP 0.906 0.906 0.700 0.700 0.700 0.864 0.864 0.864 0.564 0.812 2.000 4.400
XGBoost 0.905 0.905 0.667 0.684 0.650 0.854 0.844 0.864 0.521 0.797 1.900 4.500
KNN 0.892 0.892 0.667 0.750 0.600 0.870 0.833 0.909 0.545 0.812 1.600 4.800
LogisticRegression 0.864 0.864 0.700 0.700 0.700 0.864 0.864 0.864 0.564 0.812 2.000 4.400
NaiveBayes 0.845 0.845 0.684 0.722 0.650 0.867 0.848 0.886 0.553 0.812 1.800 4.600
DecisionTree 0.732 0.732 0.632 0.667 0.600 0.844 0.826 0.864 0.478 0.781 1.800 4.600
EEG_FAA_REL_ALPHA_F4F3 （前額葉Alpha不對稱）
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EEG_FAA_REL_ALPHA_F4F3']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
NaiveBayes 0.939 0.939 0.686 0.800 0.600 0.882 0.837 0.932 0.582 0.828 1.500 4.900
SVM 0.907 0.907 0.826 0.731 0.950 0.902 0.974 0.841 0.746 0.875 2.600 3.800
MLP 0.902 0.902 0.700 0.700 0.700 0.864 0.864 0.864 0.564 0.812 2.000 4.400
LogisticRegression 0.901 0.901 0.718 0.737 0.700 0.876 0.867 0.886 0.595 0.828 1.900 4.500
KNN 0.892 0.892 0.683 0.667 0.700 0.851 0.860 0.841 0.534 0.797 2.100 4.300
RandomForest 0.891 0.891 0.611 0.688 0.550 0.848 0.812 0.886 0.467 0.781 1.600 4.800
XGBoost 0.884 0.884 0.611 0.688 0.550 0.848 0.812 0.886 0.467 0.781 1.600 4.800
DecisionTree 0.686 0.686 0.571 0.545 0.600 0.791 0.810 0.773 0.364 0.719 2.200 4.200
EEG_FAA_ABS_ALPHA_T4T3 (顳葉Alpha不對稱)
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EEG_FAA_ABS_ALPHA_T4T3']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.924 0.924 0.765 0.929 0.650 0.915 0.860 0.977 0.703 0.875 1.400 5.000
XGBoost 0.917 0.917 0.683 0.667 0.700 0.851 0.860 0.841 0.534 0.797 2.100 4.300
RandomForest 0.908 0.908 0.789 0.833 0.750 0.911 0.891 0.932 0.703 0.875 1.800 4.600
MLP 0.908 0.908 0.700 0.700 0.700 0.864 0.864 0.864 0.564 0.812 2.000 4.400
NaiveBayes 0.908 0.908 0.667 0.750 0.600 0.870 0.833 0.909 0.545 0.812 1.600 4.800
LogisticRegression 0.898 0.898 0.737 0.778 0.700 0.889 0.870 0.909 0.628 0.844 1.800 4.600
SVM 0.870 0.870 0.650 0.650 0.650 0.841 0.841 0.841 0.491 0.781 2.000 4.400
DecisionTree 0.745 0.745 0.650 0.650 0.650 0.841 0.841 0.841 0.491 0.781 2.000 4.400
EEG_PAC_THETA_BETA2_MVL_P4
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EEG_PAC_THETA_BETA2_MVL_P4']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.959 0.959 0.857 0.818 0.900 0.930 0.952 0.909 0.790 0.906 2.200 4.200
NaiveBayes 0.950 0.950 0.722 0.812 0.650 0.891 0.854 0.932 0.623 0.844 1.600 4.800
SVM 0.950 0.950 0.905 0.864 0.950 0.953 0.976 0.932 0.861 0.938 2.200 4.200
MLP 0.944 0.944 0.750 0.750 0.750 0.886 0.886 0.886 0.636 0.844 2.000 4.400
RandomForest 0.928 0.928 0.743 0.867 0.650 0.903 0.857 0.955 0.661 0.859 1.500 4.900
LogisticRegression 0.916 0.916 0.634 0.619 0.650 0.828 0.837 0.818 0.462 0.766 2.100 4.300
XGBoost 0.911 0.911 0.769 0.789 0.750 0.899 0.889 0.909 0.669 0.859 1.900 4.500
DecisionTree 0.707 0.707 0.595 0.647 0.550 0.835 0.809 0.864 0.434 0.766 1.700 4.700
EEG_PAC_THETA_BETA2_MVL_P3 + EEG_PAC_THETA_BETA2_MVL_P4
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.966 0.966 0.800 0.800 0.800 0.909 0.909 0.909 0.709 0.875 2.000 4.400
KNN 0.953 0.953 0.780 0.762 0.800 0.897 0.907 0.886 0.678 0.859 2.100 4.300
SVM 0.947 0.947 0.850 0.850 0.850 0.932 0.932 0.932 0.782 0.906 2.000 4.400
NaiveBayes 0.932 0.932 0.789 0.833 0.750 0.911 0.891 0.932 0.703 0.875 1.800 4.600
RandomForest 0.929 0.929 0.686 0.800 0.600 0.882 0.837 0.932 0.582 0.828 1.500 4.900
LogisticRegression 0.909 0.909 0.791 0.739 0.850 0.894 0.927 0.864 0.689 0.859 2.300 4.100
XGBoost 0.907 0.907 0.780 0.762 0.800 0.897 0.907 0.886 0.678 0.859 2.100 4.300
DecisionTree 0.745 0.745 0.650 0.650 0.650 0.841 0.841 0.841 0.491 0.781 2.000 4.400
EEG_PAC_THETA_BETA2_MI_FZ
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
LogisticRegression 0.925 0.925 0.718 0.737 0.700 0.876 0.867 0.886 0.595 0.828 1.900 4.500
NaiveBayes 0.914 0.914 0.667 0.750 0.600 0.870 0.833 0.909 0.545 0.812 1.600 4.800
SVM 0.910 0.910 0.744 0.696 0.800 0.871 0.902 0.841 0.619 0.828 2.300 4.100
KNN 0.897 0.897 0.629 0.733 0.550 0.860 0.816 0.909 0.502 0.797 1.500 4.900
MLP 0.876 0.876 0.564 0.579 0.550 0.809 0.800 0.818 0.374 0.734 1.900 4.500
RandomForest 0.870 0.870 0.541 0.588 0.500 0.813 0.787 0.841 0.358 0.734 1.700 4.700
XGBoost 0.857 0.857 0.632 0.667 0.600 0.844 0.826 0.864 0.478 0.781 1.800 4.600
DecisionTree 0.684 0.684 0.564 0.579 0.550 0.809 0.800 0.818 0.374 0.734 1.900 4.500
’EEG_PAC_THETA_BETA1_MVL_F7’, ‘EEG_PAC_THETA_BETA1_MVL_F8’
[資料] 來源=isi_raw_data_transformer 目標=3TP 列數=64 特徵數=4
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EEG_PAC_THETA_BETA1_MVL_F7', 'EEG_PAC_THETA_BETA1_MVL_F8']
[CV] Stratified 10-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = weighted
[Leakage check] Class balance
count percent%
3TP
0 44 68.8
1 20 31.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 10-fold CV) ===
model AUC AUC_overall F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
SVM 0.940 0.940 0.810 0.773 0.850 0.907 0.929 0.886 0.719 0.875 2.200 4.200
XGBoost 0.940 0.940 0.810 0.773 0.850 0.907 0.929 0.886 0.719 0.875 2.200 4.200
KNN 0.929 0.929 0.743 0.867 0.650 0.903 0.857 0.955 0.661 0.859 1.500 4.900
RandomForest 0.927 0.927 0.842 0.889 0.800 0.933 0.913 0.955 0.778 0.906 1.800 4.600
MLP 0.916 0.916 0.750 0.750 0.750 0.886 0.886 0.886 0.636 0.844 2.000 4.400
NaiveBayes 0.916 0.916 0.686 0.800 0.600 0.882 0.837 0.932 0.582 0.828 1.500 4.900
LogisticRegression 0.897 0.897 0.667 0.636 0.700 0.837 0.857 0.818 0.506 0.781 2.200 4.200
DecisionTree 0.793 0.793 0.718 0.737 0.700 0.876 0.867 0.886 0.595 0.828 1.900 4.500