bai_lcga_group_2
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_2 --allow-cols BDI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_2 列數=151 特徵數=1
[特徵] 使用欄位（前 15）：['BDI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_2
0 88 58.3
1 63 41.7
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
DecisionTree 0.557 0.500 0.444 0.571 0.544 0.614 0.489 0.059 0.523 16.200 14.000
RandomForest 0.547 0.459 0.431 0.492 0.563 0.595 0.534 0.026 0.517 14.400 15.800
XGBoost 0.528 0.476 0.425 0.540 0.528 0.592 0.477 0.017 0.503 16.000 14.200
SVM 0.527 0.500 0.406 0.651 0.406 0.560 0.318 -0.033 0.457 20.200 10.000
NaiveBayes 0.454 0.031 0.500 0.016 0.734 0.584 0.989 0.019 0.583 0.400 29.800
KNN 0.451 0.322 0.345 0.302 0.565 0.542 0.591 -0.110 0.470 11.000 19.200
LogisticRegression 0.435 0.438 0.386 0.508 0.474 0.544 0.420 -0.071 0.457 16.600 13.600
MLP 0.420 0.000 0.000 0.000 0.720 0.574 0.966 -0.120 0.563 0.600 29.600
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
DecisionTree 43 45 27 36
RandomForest 47 41 32 31
XGBoost 42 46 29 34
SVM 28 60 22 41
NaiveBayes 87 1 62 1
KNN 52 36 44 19
LogisticRegression 37 51 31 32
MLP 85 3 63 0
[LOSO] skipped (--skip-loso)
IGT_NET_3
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_2 --allow-cols BDI_T1,IGT_NET_3 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'IGT_NET_3']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_2
0 88 58.3
1 63 41.7
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.595 0.492 0.478 0.508 0.616 0.631 0.602 0.109 0.563 13.400 16.800
RandomForest 0.559 0.437 0.464 0.413 0.634 0.611 0.659 0.073 0.556 11.200 19.000
MLP 0.556 0.463 0.483 0.444 0.641 0.624 0.659 0.105 0.570 11.600 18.600
DecisionTree 0.555 0.493 0.465 0.524 0.595 0.625 0.568 0.091 0.550 14.200 16.000
LogisticRegression 0.547 0.453 0.446 0.460 0.598 0.605 0.591 0.051 0.536 13.000 17.200
SVM 0.541 0.522 0.436 0.651 0.483 0.614 0.398 0.049 0.503 18.800 11.400
KNN 0.522 0.342 0.396 0.302 0.618 0.573 0.670 -0.030 0.517 9.600 20.600
NaiveBayes 0.511 0.312 0.370 0.270 0.611 0.562 0.670 -0.064 0.503 9.200 21.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
XGBoost 53 35 31 32
RandomForest 58 30 37 26
MLP 58 30 35 28
DecisionTree 50 38 30 33
LogisticRegression 52 36 34 29
SVM 35 53 22 41
KNN 59 29 44 19
NaiveBayes 59 29 46 17
[LOSO] skipped (--skip-loso)
ERQ_ES
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_2 --allow-cols BDI_T1,ERQ_ES --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'ERQ_ES']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_2
0 88 58.3
1 63 41.7
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.618 0.517 0.544 0.492 0.681 0.660 0.705 0.200 0.616 11.400 18.800
RandomForest 0.614 0.487 0.518 0.460 0.667 0.642 0.693 0.157 0.596 11.200 19.000
LogisticRegression 0.581 0.522 0.480 0.571 0.598 0.645 0.557 0.126 0.563 15.000 15.200
NaiveBayes 0.565 0.242 0.393 0.175 0.673 0.577 0.807 -0.024 0.543 5.600 24.600
DecisionTree 0.561 0.484 0.477 0.492 0.621 0.628 0.614 0.105 0.563 13.000 17.200
KNN 0.558 0.441 0.473 0.413 0.641 0.615 0.670 0.085 0.563 11.000 19.200
XGBoost 0.533 0.427 0.412 0.444 0.561 0.578 0.545 -0.010 0.503 13.600 16.600
SVM 0.533 0.500 0.427 0.603 0.493 0.597 0.420 0.024 0.497 17.800 12.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
MLP 62 26 32 31
RandomForest 61 27 34 29
LogisticRegression 49 39 27 36
NaiveBayes 71 17 52 11
DecisionTree 54 34 32 31
KNN 59 29 37 26
XGBoost 48 40 35 28
SVM 37 51 25 38
[LOSO] skipped (--skip-loso)
CPT_HRT_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_2 --allow-cols BDI_T1,CPT_HRT_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'CPT_HRT_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_2
0 88 58.3
1 63 41.7
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.638 0.520 0.516 0.524 0.651 0.655 0.648 0.171 0.596 12.800 17.400
LogisticRegression 0.568 0.517 0.452 0.603 0.542 0.627 0.477 0.080 0.530 16.800 13.400
MLP 0.562 0.472 0.469 0.476 0.617 0.621 0.614 0.090 0.556 12.800 17.400
NaiveBayes 0.558 0.250 0.440 0.175 0.692 0.587 0.841 0.021 0.563 5.000 25.200
DecisionTree 0.550 0.493 0.465 0.524 0.595 0.625 0.568 0.091 0.550 14.200 16.000
RandomForest 0.535 0.364 0.426 0.317 0.635 0.587 0.693 0.011 0.536 9.400 20.800
KNN 0.449 0.298 0.333 0.270 0.574 0.540 0.614 -0.121 0.470 10.200 20.000
SVM 0.403 0.486 0.424 0.571 0.506 0.591 0.443 0.015 0.497 17.000 13.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
XGBoost 57 31 30 33
LogisticRegression 42 46 25 38
MLP 54 34 33 30
NaiveBayes 74 14 52 11
DecisionTree 50 38 30 33
RandomForest 61 27 43 20
KNN 54 34 46 17
SVM 39 49 27 36
[LOSO] skipped (--skip-loso)
CD_RISC_T3
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_2 --allow-cols BDI_T1,CD_RISC_T3 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'CD_RISC_T3']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_2
0 88 58.3
1 63 41.7
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
LogisticRegression 0.571 0.492 0.478 0.508 0.616 0.631 0.602 0.109 0.563 13.400 16.800
MLP 0.541 0.444 0.444 0.444 0.602 0.602 0.602 0.047 0.536 12.600 17.600
RandomForest 0.534 0.397 0.414 0.381 0.597 0.581 0.614 -0.005 0.517 11.600 18.600
SVM 0.522 0.490 0.438 0.556 0.541 0.606 0.489 0.044 0.517 16.000 14.200
XGBoost 0.515 0.425 0.422 0.429 0.583 0.586 0.580 0.008 0.517 12.800 17.400
DecisionTree 0.496 0.409 0.406 0.413 0.571 0.575 0.568 -0.019 0.503 12.800 17.400
NaiveBayes 0.464 0.104 0.286 0.063 0.693 0.569 0.886 -0.085 0.543 2.800 27.400
KNN 0.454 0.355 0.361 0.349 0.551 0.544 0.557 -0.094 0.470 12.200 18.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
LogisticRegression 53 35 31 32
MLP 53 35 35 28
RandomForest 54 34 39 24
SVM 43 45 28 35
XGBoost 51 37 36 27
DecisionTree 50 38 37 26
NaiveBayes 78 10 59 4
KNN 49 39 41 22
[LOSO] skipped (--skip-loso)
HRV_PNN50
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_2 --allow-cols BDI_T1,HRV_PNN50 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'HRV_PNN50']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_2
0 88 58.3
1 63 41.7
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.565 0.485 0.452 0.524 0.578 0.615 0.545 0.068 0.536 14.600 15.600
SVM 0.565 0.364 0.348 0.381 0.506 0.524 0.489 -0.129 0.444 13.800 16.400
MLP 0.520 0.426 0.441 0.413 0.611 0.598 0.625 0.038 0.536 11.800 18.400
RandomForest 0.498 0.406 0.400 0.413 0.563 0.570 0.557 -0.030 0.497 13.000 17.200
DecisionTree 0.478 0.394 0.377 0.413 0.529 0.549 0.511 -0.075 0.470 13.800 16.400
KNN 0.471 0.342 0.370 0.317 0.584 0.557 0.614 -0.071 0.490 10.800 19.400
NaiveBayes 0.372 0.152 0.375 0.095 0.700 0.578 0.886 -0.029 0.556 3.200 27.000
LogisticRegression 0.359 0.392 0.350 0.444 0.453 0.507 0.409 -0.145 0.424 16.000 14.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
XGBoost 48 40 30 33
SVM 43 45 39 24
MLP 55 33 37 26
RandomForest 49 39 37 26
DecisionTree 45 43 37 26
KNN 54 34 43 20
NaiveBayes 78 10 57 6
LogisticRegression 36 52 35 28
[LOSO] skipped (--skip-loso)
bai_lcga_group_3
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_3 --allow-cols BDI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_3 列數=151 特徵數=1
[特徵] 使用欄位（前 15）：['BDI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_3
0 62 41.1
1 81 53.6
2 8 5.3
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.528 0.463 0.458 0.470 0.318 0.316 0.321 0.001 0.470 17.600 11.400
LogisticRegression 0.522 0.398 0.444 0.417 0.309 0.331 0.417 0.046 0.417 18.400 3.800
NaiveBayes 0.520 0.354 0.280 0.483 0.220 0.174 0.300 -0.117 0.483 28.000 1.200
RandomForest 0.514 0.402 0.429 0.384 0.282 0.297 0.273 -0.054 0.384 12.400 13.800
DecisionTree 0.512 0.392 0.430 0.371 0.277 0.298 0.266 -0.051 0.371 11.200 14.200
KNN 0.511 0.520 0.505 0.536 0.360 0.352 0.370 0.098 0.536 18.200 12.000
MLP 0.498 0.500 0.526 0.563 0.336 0.369 0.367 0.123 0.563 25.000 5.200
SVM 0.412 0.339 0.368 0.377 0.245 0.260 0.348 0.006 0.377 19.200 1.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 23 36 3
True_1 30 48 3
True_2 4 4 0
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 6 37 19
True_1 11 53 17
True_2 2 2 4
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 0 60 2
True_1 5 73 3
True_2 1 7 0
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 27 27 8
True_1 38 31 12
True_2 4 4 0
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 28 24 10
True_1 39 28 14
True_2 4 4 0
[KNN]
Pred_0 Pred_1 Pred_2
True_0 29 33 0
True_1 29 52 0
True_2 2 6 0
[MLP]
Pred_0 Pred_1 Pred_2
True_0 14 48 0
True_1 10 71 0
True_2 2 6 0
[SVM]
Pred_0 Pred_1 Pred_2
True_0 1 39 22
True_1 4 53 24
True_2 1 4 3
[LOSO] skipped (--skip-loso)
CD_RISC_T1
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_3 --allow-cols BDI_T1,CD_RISC_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'CD_RISC_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_3
0 62 41.1
1 81 53.6
2 8 5.3
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
LogisticRegression 0.604 0.526 0.580 0.497 0.418 0.437 0.489 0.187 0.497 14.200 9.000
NaiveBayes 0.582 0.494 0.529 0.536 0.372 0.414 0.389 0.100 0.536 23.600 5.200
SVM 0.544 0.500 0.524 0.483 0.363 0.376 0.367 0.103 0.483 16.000 9.800
XGBoost 0.538 0.474 0.465 0.483 0.327 0.322 0.333 0.014 0.483 17.400 12.200
KNN 0.512 0.475 0.461 0.490 0.328 0.320 0.337 0.010 0.490 17.800 12.400
MLP 0.509 0.457 0.451 0.464 0.315 0.312 0.318 -0.013 0.464 17.200 12.000
RandomForest 0.503 0.462 0.449 0.477 0.319 0.310 0.328 -0.016 0.477 17.800 12.400
DecisionTree 0.489 0.428 0.442 0.424 0.302 0.307 0.302 -0.027 0.424 13.000 15.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 24 22 16
True_1 19 47 15
True_2 2 2 4
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 14 46 2
True_1 11 66 4
True_2 1 6 1
[SVM]
Pred_0 Pred_1 Pred_2
True_0 23 27 12
True_1 23 49 9
True_2 3 4 1
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 26 35 1
True_1 32 47 2
True_2 3 5 0
[KNN]
Pred_0 Pred_1 Pred_2
True_0 26 36 0
True_1 33 48 0
True_2 3 5 0
[MLP]
Pred_0 Pred_1 Pred_2
True_0 24 35 3
True_1 33 46 2
True_2 3 5 0
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 25 37 0
True_1 34 47 0
True_2 3 5 0
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 31 28 3
True_1 40 33 8
True_2 4 4 0
[LOSO] skipped (--skip-loso)
ERQ_ES
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_3 --allow-cols BDI_T1,ERQ_ES --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'ERQ_ES']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_3
0 62 41.1
1 81 53.6
2 8 5.3
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.615 0.549 0.551 0.550 0.474 0.501 0.460 0.162 0.550 15.400 13.800
RandomForest 0.575 0.545 0.538 0.556 0.383 0.375 0.394 0.165 0.556 15.200 14.600
LogisticRegression 0.570 0.535 0.573 0.517 0.450 0.448 0.542 0.197 0.517 14.400 10.000
XGBoost 0.561 0.501 0.493 0.510 0.349 0.343 0.355 0.069 0.510 16.800 12.800
NaiveBayes 0.542 0.486 0.503 0.523 0.406 0.435 0.418 0.075 0.523 22.800 6.000
KNN 0.532 0.526 0.546 0.530 0.482 0.676 0.440 0.098 0.530 17.600 12.200
DecisionTree 0.526 0.474 0.484 0.477 0.335 0.337 0.342 0.053 0.477 13.000 15.800
SVM 0.446 0.481 0.500 0.470 0.394 0.389 0.435 0.079 0.470 15.400 10.600
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[MLP]
Pred_0 Pred_1 Pred_2
True_0 34 27 1
True_1 32 47 2
True_2 3 3 2
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 38 24 0
True_1 33 46 2
True_2 2 6 0
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 26 25 11
True_1 21 47 13
True_2 3 0 5
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 30 31 1
True_1 32 47 2
True_2 2 6 0
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 14 46 2
True_1 15 63 3
True_2 1 5 2
[KNN]
Pred_0 Pred_1 Pred_2
True_0 28 34 0
True_1 31 50 0
True_2 2 4 2
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 36 24 2
True_1 40 36 5
True_2 3 5 0
[SVM]
Pred_0 Pred_1 Pred_2
True_0 24 31 7
True_1 26 44 11
True_2 3 2 3
[LOSO] skipped (--skip-loso)
HRV_NN50
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_3 --allow-cols BDI_T1,HRV_NN50 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'HRV_NN50']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_3
0 62 41.1
1 81 53.6
2 8 5.3
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
LogisticRegression 0.516 0.405 0.448 0.391 0.310 0.329 0.371 0.009 0.391 15.600 7.000
SVM 0.485 0.389 0.424 0.377 0.287 0.306 0.324 -0.033 0.377 16.400 6.800
XGBoost 0.480 0.452 0.447 0.457 0.315 0.312 0.318 -0.021 0.457 16.600 12.600
NaiveBayes 0.470 0.365 0.297 0.477 0.307 0.271 0.371 -0.082 0.477 26.600 2.200
RandomForest 0.449 0.433 0.430 0.437 0.340 0.347 0.337 -0.060 0.437 17.400 11.600
MLP 0.448 0.402 0.395 0.411 0.275 0.273 0.279 -0.123 0.411 18.200 11.000
KNN 0.445 0.449 0.435 0.477 0.305 0.300 0.319 -0.040 0.477 20.600 9.600
DecisionTree 0.441 0.380 0.383 0.377 0.296 0.294 0.299 -0.136 0.377 15.600 12.600
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 12 32 18
True_1 20 44 17
True_2 3 2 3
[SVM]
Pred_0 Pred_1 Pred_2
True_0 11 36 15
True_1 19 44 18
True_2 4 2 2
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 27 34 1
True_1 35 42 4
True_2 1 7 0
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 0 58 4
True_1 10 70 1
True_2 1 5 2
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 22 39 1
True_1 34 43 4
True_2 2 5 1
[MLP]
Pred_0 Pred_1 Pred_2
True_0 19 42 1
True_1 34 43 4
True_2 2 6 0
[KNN]
Pred_0 Pred_1 Pred_2
True_0 18 44 0
True_1 27 54 0
True_2 3 5 0
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 21 37 4
True_1 41 35 5
True_2 1 6 1
[LOSO] skipped (--skip-loso)
IGT_NET_3
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_3 --allow-cols BDI_T1,IGT_NET_3 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'IGT_NET_3']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_3
0 62 41.1
1 81 53.6
2 8 5.3
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.542 0.488 0.479 0.497 0.338 0.332 0.344 0.043 0.497 17.000 12.600
SVM 0.534 0.445 0.490 0.430 0.349 0.369 0.402 0.047 0.430 16.600 7.200
DecisionTree 0.532 0.488 0.493 0.483 0.367 0.367 0.371 0.061 0.483 16.200 11.600
LogisticRegression 0.525 0.394 0.446 0.377 0.345 0.351 0.483 0.019 0.377 12.600 9.400
KNN 0.519 0.475 0.461 0.490 0.328 0.320 0.337 0.010 0.490 17.800 12.400
MLP 0.499 0.428 0.426 0.430 0.341 0.353 0.334 -0.069 0.430 16.400 12.800
NaiveBayes 0.485 0.425 0.434 0.497 0.332 0.394 0.355 -0.030 0.497 25.200 4.400
RandomForest 0.446 0.368 0.358 0.384 0.247 0.243 0.255 -0.185 0.384 19.200 10.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 28 33 1
True_1 32 47 2
True_2 3 5 0
[SVM]
Pred_0 Pred_1 Pred_2
True_0 17 35 10
True_1 17 45 19
True_2 2 3 3
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 26 31 5
True_1 29 46 6
True_2 3 4 1
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 18 29 15
True_1 28 33 20
True_2 1 1 6
[KNN]
Pred_0 Pred_1 Pred_2
True_0 26 36 0
True_1 33 48 0
True_2 3 5 0
[MLP]
Pred_0 Pred_1 Pred_2
True_0 23 37 2
True_1 38 41 2
True_2 3 4 1
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 7 54 1
True_1 13 67 1
True_2 2 5 1
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 13 48 1
True_1 32 45 4
True_2 5 3 0
[LOSO] skipped (--skip-loso)
ACS_FOCUSING
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bai_lcga_group_3 --allow-cols BDI_T1,ACS_FOCUSING --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bai_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'ACS_FOCUSING']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bai_lcga_group_3
0 62 41.1
1 81 53.6
2 8 5.3
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.562 0.511 0.500 0.523 0.354 0.347 0.362 0.086 0.523 17.400 12.400
LogisticRegression 0.508 0.327 0.358 0.318 0.251 0.260 0.319 -0.100 0.318 15.000 7.000
SVM 0.499 0.383 0.412 0.371 0.322 0.321 0.401 -0.045 0.371 14.400 9.600
DecisionTree 0.462 0.419 0.422 0.417 0.368 0.375 0.363 -0.074 0.417 15.200 13.600
NaiveBayes 0.449 0.395 0.390 0.483 0.253 0.262 0.305 -0.055 0.483 25.800 3.200
RandomForest 0.420 0.365 0.359 0.371 0.251 0.246 0.256 -0.195 0.371 16.200 13.600
MLP 0.419 0.396 0.397 0.397 0.323 0.347 0.312 -0.131 0.397 15.800 13.600
KNN 0.383 0.394 0.384 0.404 0.270 0.263 0.278 -0.148 0.404 16.800 13.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 29 32 1
True_1 30 50 1
True_2 3 5 0
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 7 34 21
True_1 26 38 17
True_2 2 3 3
[SVM]
Pred_0 Pred_1 Pred_2
True_0 16 33 13
True_1 31 36 14
True_2 1 3 4
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 23 35 4
True_1 42 38 1
True_2 3 3 2
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 4 55 3
True_1 9 69 3
True_2 3 5 0
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 20 40 2
True_1 45 36 0
True_2 3 5 0
[MLP]
Pred_0 Pred_1 Pred_2
True_0 22 38 2
True_1 43 37 1
True_2 3 4 1
[KNN]
Pred_0 Pred_1 Pred_2
True_0 21 41 0
True_1 41 40 0
True_2 5 3 0
[LOSO] skipped (--skip-loso)
bdi_lcga_group_2
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_2 --allow-cols BAI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_2 列數=151 特徵數=1
[特徵] 使用欄位（前 15）：['BAI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_2
0 53 35.1
1 98 64.9
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
SVM 0.579 0.622 0.622 0.622 0.302 0.302 0.302 -0.076 0.510 19.600 10.600
KNN 0.481 0.660 0.638 0.684 0.303 0.326 0.283 -0.035 0.543 21.000 9.200
RandomForest 0.459 0.608 0.615 0.602 0.296 0.291 0.302 -0.095 0.497 19.200 11.000
DecisionTree 0.457 0.589 0.609 0.571 0.304 0.288 0.321 -0.105 0.483 18.400 11.800
XGBoost 0.448 0.573 0.609 0.541 0.325 0.297 0.358 -0.097 0.477 17.400 12.800
LogisticRegression 0.418 0.541 0.590 0.500 0.314 0.279 0.358 -0.136 0.450 16.600 13.600
NaiveBayes 0.393 0.777 0.644 0.980 0.000 0.000 0.000 -0.085 0.636 29.800 0.400
MLP 0.390 0.787 0.649 1.000 0.000 0.000 0.000 0.000 0.649 30.200 0.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
SVM 16 37 37 61
KNN 15 38 31 67
RandomForest 16 37 39 59
DecisionTree 17 36 42 56
XGBoost 19 34 45 53
LogisticRegression 19 34 49 49
NaiveBayes 0 53 2 96
MLP 0 53 0 98
[LOSO] skipped (--skip-loso)
WCST_TOTAL_ERRORS_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_2 --allow-cols BAI_T1,WCST_TOTAL_ERRORS_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'WCST_TOTAL_ERRORS_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_2
0 53 35.1
1 98 64.9
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
SVM 0.510 0.631 0.602 0.663 0.208 0.233 0.189 -0.157 0.497 21.600 8.600
NaiveBayes 0.490 0.759 0.657 0.898 0.200 0.412 0.132 0.045 0.629 26.800 3.400
DecisionTree 0.485 0.636 0.639 0.633 0.336 0.333 0.340 -0.028 0.530 19.400 10.800
RandomForest 0.480 0.679 0.640 0.724 0.280 0.325 0.245 -0.033 0.556 22.200 8.000
LogisticRegression 0.468 0.614 0.637 0.592 0.354 0.333 0.377 -0.030 0.517 18.200 12.000
XGBoost 0.463 0.598 0.640 0.561 0.373 0.338 0.415 -0.023 0.510 17.200 13.000
KNN 0.454 0.697 0.633 0.776 0.214 0.290 0.170 -0.065 0.563 24.000 6.200
MLP 0.383 0.610 0.598 0.622 0.235 0.245 0.226 -0.154 0.483 20.400 9.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
SVM 10 43 33 65
NaiveBayes 7 46 10 88
DecisionTree 18 35 36 62
RandomForest 13 40 27 71
LogisticRegression 20 33 40 58
XGBoost 22 31 43 55
KNN 9 44 22 76
MLP 12 41 37 61
[LOSO] skipped (--skip-loso)
CPT_COMMISSION_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_2 --allow-cols BAI_T1,CPT_COMMISSION_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'CPT_COMMISSION_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_2
0 53 35.1
1 98 64.9
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
DecisionTree 0.506 0.639 0.656 0.622 0.378 0.362 0.396 0.018 0.543 18.600 11.600
XGBoost 0.480 0.646 0.640 0.653 0.327 0.333 0.321 -0.026 0.536 20.000 10.200
RandomForest 0.479 0.646 0.640 0.653 0.327 0.333 0.321 -0.026 0.536 20.000 10.200
MLP 0.455 0.650 0.637 0.663 0.314 0.327 0.302 -0.036 0.536 20.400 9.800
SVM 0.440 0.459 0.610 0.367 0.414 0.326 0.566 -0.065 0.437 11.800 18.400
KNN 0.439 0.686 0.643 0.735 0.283 0.333 0.245 -0.022 0.563 22.400 7.800
NaiveBayes 0.428 0.777 0.644 0.980 0.000 0.000 0.000 -0.085 0.636 29.800 0.400
LogisticRegression 0.365 0.497 0.557 0.449 0.288 0.250 0.340 -0.202 0.411 15.800 14.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
DecisionTree 21 32 37 61
XGBoost 17 36 34 64
RandomForest 17 36 34 64
MLP 16 37 33 65
SVM 30 23 62 36
KNN 13 40 26 72
NaiveBayes 0 53 2 96
LogisticRegression 18 35 54 44
[LOSO] skipped (--skip-loso)
HRV_NN50
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_2 --allow-cols BAI_T1,HRV_NN50 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'HRV_NN50']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_2
0 53 35.1
1 98 64.9
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
SVM 0.575 0.317 0.537 0.224 0.417 0.309 0.642 -0.144 0.371 8.200 22.000
LogisticRegression 0.538 0.537 0.694 0.439 0.479 0.382 0.642 0.078 0.510 12.400 17.800
XGBoost 0.486 0.615 0.667 0.571 0.417 0.373 0.472 0.041 0.536 16.800 13.400
NaiveBayes 0.468 0.757 0.634 0.939 0.000 0.000 0.000 -0.150 0.609 29.000 1.200
RandomForest 0.454 0.636 0.630 0.643 0.308 0.314 0.302 -0.056 0.523 20.000 10.200
DecisionTree 0.449 0.611 0.630 0.592 0.339 0.322 0.358 -0.049 0.510 18.400 11.800
KNN 0.448 0.682 0.637 0.735 0.264 0.316 0.226 -0.043 0.556 22.600 7.600
MLP 0.406 0.616 0.610 0.622 0.269 0.275 0.264 -0.114 0.497 20.000 10.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
SVM 34 19 76 22
LogisticRegression 34 19 55 43
XGBoost 25 28 42 56
NaiveBayes 0 53 6 92
RandomForest 16 37 35 63
DecisionTree 19 34 40 58
KNN 12 41 26 72
MLP 14 39 37 61
[LOSO] skipped (--skip-loso)
HRV_SDNN_MS
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_2 --allow-cols BAI_T1,HRV_SDNN_MS --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'HRV_SDNN_MS']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_2
0 53 35.1
1 98 64.9
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
SVM 0.577 0.506 0.579 0.449 0.328 0.280 0.396 -0.148 0.430 15.200 15.000
DecisionTree 0.529 0.674 0.684 0.663 0.422 0.411 0.434 0.096 0.583 19.000 11.200
KNN 0.483 0.686 0.651 0.724 0.316 0.357 0.283 0.008 0.570 21.800 8.400
LogisticRegression 0.483 0.579 0.624 0.541 0.353 0.318 0.396 -0.061 0.490 17.000 13.200
RandomForest 0.478 0.694 0.694 0.694 0.434 0.434 0.434 0.128 0.603 19.600 10.600
XGBoost 0.457 0.587 0.628 0.551 0.356 0.323 0.396 -0.051 0.497 17.200 13.000
MLP 0.432 0.653 0.635 0.673 0.300 0.319 0.283 -0.045 0.536 20.800 9.400
NaiveBayes 0.401 0.760 0.639 0.939 0.033 0.143 0.019 -0.096 0.616 28.800 1.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
SVM 21 32 54 44
DecisionTree 23 30 33 65
KNN 15 38 27 71
LogisticRegression 21 32 45 53
RandomForest 23 30 30 68
XGBoost 21 32 44 54
MLP 15 38 32 66
NaiveBayes 1 52 6 92
[LOSO] skipped (--skip-loso)
CPT_DPR_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_2 --allow-cols BAI_T1,CPT_DPR_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'CPT_DPR_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_2
0 53 35.1
1 98 64.9
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
SVM 0.623 0.467 0.512 0.429 0.213 0.188 0.245 -0.312 0.364 16.400 13.800
DecisionTree 0.461 0.585 0.611 0.561 0.316 0.295 0.340 -0.096 0.483 18.000 12.200
KNN 0.458 0.705 0.670 0.745 0.358 0.405 0.321 0.070 0.596 21.800 8.400
MLP 0.432 0.615 0.628 0.602 0.327 0.316 0.340 -0.057 0.510 18.800 11.400
RandomForest 0.398 0.612 0.612 0.612 0.283 0.283 0.283 -0.105 0.497 19.600 10.600
XGBoost 0.358 0.527 0.571 0.490 0.283 0.254 0.321 -0.182 0.430 16.800 13.400
LogisticRegression 0.341 0.503 0.541 0.469 0.235 0.212 0.264 -0.256 0.397 17.000 13.200
NaiveBayes 0.323 0.767 0.639 0.959 0.000 0.000 0.000 -0.121 0.623 29.400 0.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
SVM 13 40 56 42
DecisionTree 18 35 43 55
KNN 17 36 25 73
MLP 18 35 39 59
RandomForest 15 38 38 60
XGBoost 17 36 50 48
LogisticRegression 14 39 52 46
NaiveBayes 0 53 4 94
[LOSO] skipped (--skip-loso)
bdi_lcga_group_3
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_3 --allow-cols BAI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_3 列數=151 特徵數=1
[特徵] 使用欄位（前 15）：['BAI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_3
0 53 35.1
1 88 58.3
2 10 6.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
LogisticRegression 0.504 0.371 0.402 0.371 0.289 0.289 0.399 -0.026 0.371 15.800 5.200
MLP 0.495 0.547 0.606 0.609 0.405 0.708 0.406 0.171 0.609 25.600 4.400
NaiveBayes 0.472 0.421 0.341 0.550 0.241 0.195 0.314 -0.048 0.550 28.400 1.000
DecisionTree 0.467 0.364 0.414 0.338 0.286 0.302 0.322 -0.066 0.338 12.400 10.400
KNN 0.441 0.481 0.459 0.517 0.315 0.308 0.330 -0.007 0.517 22.000 8.200
RandomForest 0.436 0.371 0.395 0.358 0.301 0.299 0.331 -0.102 0.358 14.600 10.600
XGBoost 0.435 0.506 0.500 0.543 0.432 0.478 0.424 0.060 0.543 23.000 6.000
SVM 0.412 0.398 0.362 0.450 0.278 0.239 0.376 0.016 0.450 21.000 1.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 4 31 18
True_1 20 46 22
True_2 2 2 6
[MLP]
Pred_0 Pred_1 Pred_2
True_0 11 42 0
True_1 8 80 0
True_2 3 6 1
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 0 50 3
True_1 4 83 1
True_2 1 9 0
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 16 25 12
True_1 34 32 22
True_2 2 5 3
[KNN]
Pred_0 Pred_1 Pred_2
True_0 14 39 0
True_1 24 64 0
True_2 3 7 0
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 15 32 6
True_1 36 36 16
True_2 2 5 3
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 10 41 2
True_1 18 69 1
True_2 2 5 3
[SVM]
Pred_0 Pred_1 Pred_2
True_0 0 37 16
True_1 7 64 17
True_2 2 4 4
[LOSO] skipped (--skip-loso)
HRV_LF
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_3 --allow-cols BAI_T1,HRV_LF --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'HRV_LF']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_3
0 53 35.1
1 88 58.3
2 10 6.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.548 0.515 0.519 0.536 0.390 0.495 0.379 0.061 0.536 20.800 9.000
NaiveBayes 0.526 0.429 0.381 0.543 0.250 0.233 0.313 -0.042 0.543 27.600 1.800
LogisticRegression 0.523 0.447 0.496 0.424 0.359 0.371 0.428 0.062 0.424 14.400 8.000
XGBoost 0.521 0.506 0.497 0.523 0.369 0.377 0.369 0.058 0.523 20.400 8.400
DecisionTree 0.493 0.450 0.461 0.444 0.365 0.371 0.363 -0.019 0.444 15.600 12.800
RandomForest 0.485 0.458 0.443 0.477 0.299 0.293 0.308 -0.040 0.477 20.000 9.400
MLP 0.463 0.463 0.452 0.483 0.370 0.393 0.363 -0.032 0.483 20.800 8.200
SVM 0.417 0.384 0.449 0.358 0.313 0.330 0.368 -0.016 0.358 11.200 11.600
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[KNN]
Pred_0 Pred_1 Pred_2
True_0 17 36 0
True_1 24 63 1
True_2 4 5 1
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 1 49 3
True_1 6 81 1
True_2 2 8 0
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 15 26 12
True_1 22 44 22
True_2 3 2 5
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 16 34 3
True_1 23 62 3
True_2 3 6 1
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 20 30 3
True_1 39 45 4
True_2 5 3 2
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 14 38 1
True_1 27 58 3
True_2 6 4 0
[MLP]
Pred_0 Pred_1 Pred_2
True_0 11 39 3
True_1 27 60 1
True_2 3 5 2
[SVM]
Pred_0 Pred_1 Pred_2
True_0 18 20 15
True_1 38 32 18
True_2 2 4 4
[LOSO] skipped (--skip-loso)
HRV_LF_HF
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_3 --allow-cols BAI_T1,HRV_LF_HF --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'HRV_LF_HF']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_3
0 53 35.1
1 88 58.3
2 10 6.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
NaiveBayes 0.520 0.440 0.406 0.550 0.260 0.256 0.319 -0.032 0.550 27.600 2.200
XGBoost 0.511 0.471 0.467 0.477 0.378 0.402 0.367 -0.011 0.477 18.600 10.400
LogisticRegression 0.505 0.419 0.467 0.397 0.336 0.347 0.408 0.022 0.397 14.200 8.000
DecisionTree 0.492 0.441 0.455 0.430 0.319 0.322 0.321 -0.024 0.430 15.800 11.400
KNN 0.480 0.482 0.461 0.510 0.318 0.310 0.332 -0.005 0.510 21.000 9.200
MLP 0.450 0.446 0.436 0.457 0.295 0.292 0.299 -0.058 0.457 19.400 9.400
RandomForest 0.447 0.449 0.437 0.464 0.296 0.291 0.303 -0.056 0.464 19.600 9.600
SVM 0.384 0.354 0.429 0.325 0.280 0.307 0.319 -0.053 0.325 10.200 12.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 2 49 2
True_1 7 81 0
True_2 2 8 0
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 14 35 4
True_1 32 56 0
True_2 6 2 2
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 13 26 14
True_1 25 42 21
True_2 2 3 5
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 18 30 5
True_1 33 46 9
True_2 6 3 1
[KNN]
Pred_0 Pred_1 Pred_2
True_0 16 37 0
True_1 27 61 0
True_2 3 7 0
[MLP]
Pred_0 Pred_1 Pred_2
True_0 15 36 2
True_1 29 54 5
True_2 3 7 0
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 15 36 2
True_1 30 55 3
True_2 3 7 0
[SVM]
Pred_0 Pred_1 Pred_2
True_0 18 19 16
True_1 40 28 20
True_2 3 4 3
[LOSO] skipped (--skip-loso)
HRV_NN50
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_3 --allow-cols BAI_T1,HRV_NN50 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'HRV_NN50']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_3
0 53 35.1
1 88 58.3
2 10 6.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
LogisticRegression 0.562 0.421 0.488 0.404 0.366 0.374 0.439 0.052 0.404 10.600 13.400
XGBoost 0.505 0.497 0.490 0.510 0.398 0.422 0.388 0.036 0.510 19.800 9.200
RandomForest 0.500 0.452 0.442 0.464 0.295 0.290 0.300 -0.043 0.464 19.000 10.000
NaiveBayes 0.499 0.459 0.444 0.517 0.290 0.296 0.315 -0.043 0.517 24.600 5.000
MLP 0.474 0.473 0.461 0.490 0.309 0.306 0.315 -0.002 0.490 20.000 8.800
DecisionTree 0.467 0.437 0.431 0.444 0.289 0.284 0.294 -0.069 0.444 18.000 11.200
KNN 0.436 0.410 0.389 0.437 0.264 0.254 0.278 -0.158 0.437 21.000 9.200
SVM 0.412 0.277 0.385 0.265 0.227 0.272 0.266 -0.107 0.265 6.600 15.600
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 24 19 10
True_1 40 32 16
True_2 3 2 5
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 15 35 3
True_1 27 60 1
True_2 4 4 2
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 14 35 4
True_1 30 56 2
True_2 6 4 0
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 8 44 1
True_1 16 70 2
True_2 1 9 0
[MLP]
Pred_0 Pred_1 Pred_2
True_0 14 34 5
True_1 26 60 2
True_2 4 6 0
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 16 35 2
True_1 34 51 3
True_2 6 4 0
[KNN]
Pred_0 Pred_1 Pred_2
True_0 11 42 0
True_1 33 55 0
True_2 2 8 0
[SVM]
Pred_0 Pred_1 Pred_2
True_0 22 13 18
True_1 52 16 20
True_2 4 4 2
[LOSO] skipped (--skip-loso)
WCST_TOTAL_ERRORS_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_3 --allow-cols BAI_T1,WCST_TOTAL_ERRORS_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'WCST_TOTAL_ERRORS_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_3
0 53 35.1
1 88 58.3
2 10 6.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
NaiveBayes 0.522 0.475 0.471 0.556 0.295 0.316 0.333 0.028 0.556 26.200 3.400
XGBoost 0.493 0.488 0.477 0.510 0.358 0.377 0.356 0.015 0.510 21.000 8.200
LogisticRegression 0.486 0.414 0.464 0.397 0.316 0.334 0.395 0.029 0.397 15.200 5.800
KNN 0.485 0.498 0.497 0.530 0.406 0.479 0.392 0.028 0.530 22.600 6.800
DecisionTree 0.478 0.446 0.449 0.444 0.332 0.336 0.331 -0.037 0.444 16.600 12.000
MLP 0.468 0.456 0.444 0.477 0.332 0.342 0.332 -0.046 0.477 20.800 8.200
RandomForest 0.450 0.446 0.441 0.464 0.332 0.384 0.325 -0.075 0.464 20.200 9.400
SVM 0.370 0.469 0.516 0.464 0.353 0.382 0.408 0.081 0.464 18.000 4.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 6 45 2
True_1 9 78 1
True_2 2 8 0
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 14 37 2
True_1 24 62 2
True_2 3 6 1
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 8 27 18
True_1 18 47 23
True_2 3 2 5
[KNN]
Pred_0 Pred_1 Pred_2
True_0 12 41 0
True_1 20 66 2
True_2 2 6 2
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 19 31 3
True_1 37 47 4
True_2 4 5 1
[MLP]
Pred_0 Pred_1 Pred_2
True_0 12 39 2
True_1 26 59 3
True_2 3 6 1
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 12 39 2
True_1 31 57 0
True_2 4 5 1
[SVM]
Pred_0 Pred_1 Pred_2
True_0 10 30 13
True_1 12 56 20
True_2 2 4 4
[LOSO] skipped (--skip-loso)
CPT_REACTION_TIME_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target bdi_lcga_group_3 --allow-cols BAI_T1,CPT_REACTION_TIME_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=bdi_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BAI_T1', 'CPT_REACTION_TIME_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
bdi_lcga_group_3
0 53 35.1
1 88 58.3
2 10 6.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.535 0.491 0.484 0.503 0.388 0.396 0.384 0.030 0.503 19.800 8.800
NaiveBayes 0.533 0.427 0.379 0.543 0.249 0.231 0.313 -0.056 0.543 27.800 1.800
LogisticRegression 0.529 0.425 0.462 0.417 0.347 0.351 0.441 0.030 0.417 16.000 6.400
DecisionTree 0.524 0.483 0.491 0.477 0.406 0.402 0.414 0.047 0.477 16.000 11.800
KNN 0.522 0.501 0.508 0.517 0.421 0.535 0.395 0.027 0.517 20.600 9.000
RandomForest 0.514 0.462 0.456 0.470 0.373 0.387 0.366 -0.029 0.470 19.200 9.600
MLP 0.476 0.481 0.478 0.483 0.383 0.386 0.381 0.019 0.483 18.200 10.200
SVM 0.422 0.377 0.467 0.364 0.325 0.350 0.392 0.013 0.364 8.600 15.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 15 34 4
True_1 27 59 2
True_2 2 6 2
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 1 50 2
True_1 6 81 1
True_2 2 8 0
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 10 31 12
True_1 20 47 21
True_2 2 2 6
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 21 27 5
True_1 36 48 4
True_2 2 5 3
[KNN]
Pred_0 Pred_1 Pred_2
True_0 16 37 0
True_1 27 60 1
True_2 2 6 2
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 15 35 3
True_1 32 54 2
True_2 1 7 2
[MLP]
Pred_0 Pred_1 Pred_2
True_0 18 32 3
True_1 31 53 4
True_2 2 6 2
[SVM]
Pred_0 Pred_1 Pred_2
True_0 26 15 12
True_1 47 25 16
True_2 3 3 4
[LOSO] skipped (--skip-loso)
cdrisc_lcga_group_2
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_2 --allow-cols BAI_T1,BDI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_2
0 22 14.6
1 129 85.4
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.635 0.913 0.857 0.977 0.077 0.250 0.045 0.049 0.841 29.400 0.800
MLP 0.592 0.874 0.864 0.884 0.195 0.211 0.182 0.070 0.781 26.400 3.800
RandomForest 0.577 0.877 0.870 0.884 0.238 0.250 0.227 0.116 0.788 26.200 4.000
DecisionTree 0.576 0.876 0.876 0.876 0.273 0.273 0.273 0.149 0.788 25.800 4.400
SVM 0.510 0.727 0.879 0.620 0.268 0.183 0.500 0.087 0.603 18.200 12.000
XGBoost 0.492 0.813 0.875 0.760 0.262 0.205 0.364 0.099 0.702 22.400 7.800
LogisticRegression 0.489 0.759 0.854 0.682 0.200 0.146 0.318 0.000 0.629 20.600 9.600
NaiveBayes 0.360 0.902 0.849 0.961 0.000 0.000 0.000 -0.076 0.821 29.200 1.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
KNN 1 21 3 126
MLP 4 18 15 114
RandomForest 5 17 15 114
DecisionTree 6 16 16 113
SVM 11 11 49 80
XGBoost 8 14 31 98
LogisticRegression 7 15 41 88
NaiveBayes 0 22 5 124
[LOSO] skipped (--skip-loso)
IGT_NET_4
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_2 --allow-cols BAI_T1,BDI_T1,IGT_NET_4 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'IGT_NET_4']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_2
0 22 14.6
1 129 85.4
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
RandomForest 0.609 0.908 0.861 0.961 0.138 0.286 0.091 0.088 0.834 28.800 1.400
XGBoost 0.607 0.812 0.882 0.752 0.286 0.220 0.409 0.128 0.702 22.000 8.200
DecisionTree 0.567 0.890 0.873 0.907 0.256 0.294 0.227 0.150 0.808 26.800 3.400
SVM 0.558 0.784 0.883 0.705 0.286 0.208 0.455 0.121 0.669 20.600 9.600
LogisticRegression 0.557 0.742 0.891 0.636 0.296 0.203 0.545 0.131 0.623 18.400 11.800
KNN 0.513 0.893 0.852 0.938 0.065 0.111 0.045 -0.025 0.808 28.400 1.800
MLP 0.511 0.875 0.853 0.899 0.108 0.133 0.091 -0.012 0.781 27.200 3.000
NaiveBayes 0.470 0.906 0.850 0.969 0.000 0.000 0.000 -0.068 0.828 29.400 0.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
RandomForest 2 20 5 124
XGBoost 9 13 32 97
DecisionTree 5 17 12 117
SVM 10 12 38 91
LogisticRegression 12 10 47 82
KNN 1 21 8 121
MLP 2 20 13 116
NaiveBayes 0 22 4 125
[LOSO] skipped (--skip-loso)
CPT_DPR_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_2 --allow-cols BAI_T1,BDI_T1,CPT_DPR_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'CPT_DPR_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_2
0 22 14.6
1 129 85.4
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
RandomForest 0.547 0.913 0.857 0.977 0.077 0.250 0.045 0.049 0.841 29.400 0.800
DecisionTree 0.540 0.859 0.866 0.853 0.217 0.208 0.227 0.077 0.762 25.400 4.800
XGBoost 0.524 0.776 0.852 0.713 0.185 0.140 0.273 -0.011 0.649 21.600 8.600
MLP 0.524 0.889 0.879 0.899 0.293 0.316 0.273 0.183 0.808 26.400 3.800
KNN 0.518 0.905 0.855 0.961 0.071 0.167 0.045 0.012 0.828 29.000 1.200
SVM 0.469 0.685 0.851 0.574 0.209 0.141 0.409 -0.012 0.550 17.400 12.800
LogisticRegression 0.424 0.723 0.853 0.628 0.205 0.143 0.364 -0.006 0.589 19.000 11.200
NaiveBayes 0.333 0.898 0.848 0.953 0.000 0.000 0.000 -0.084 0.815 29.000 1.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
RandomForest 1 21 3 126
DecisionTree 5 17 19 110
XGBoost 6 16 37 92
MLP 6 16 13 116
KNN 1 21 5 124
SVM 9 13 55 74
LogisticRegression 8 14 48 81
NaiveBayes 0 22 6 123
[LOSO] skipped (--skip-loso)
EDU
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_2 --allow-cols BAI_T1,BDI_T1,EDU --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['EDU', 'BDI_T1', 'BAI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_2
0 22 14.6
1 129 85.4
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.740 0.909 0.889 0.930 0.368 0.438 0.318 0.285 0.841 27.000 3.200
KNN 0.731 0.904 0.865 0.946 0.188 0.300 0.136 0.116 0.828 28.200 2.000
RandomForest 0.683 0.902 0.876 0.930 0.278 0.357 0.227 0.192 0.828 27.400 2.800
SVM 0.602 0.640 0.878 0.504 0.263 0.169 0.591 0.067 0.517 14.800 15.400
DecisionTree 0.594 0.875 0.882 0.868 0.304 0.292 0.318 0.180 0.788 25.400 4.800
LogisticRegression 0.572 0.595 0.879 0.450 0.262 0.165 0.636 0.061 0.477 13.200 17.000
XGBoost 0.526 0.847 0.882 0.814 0.296 0.250 0.364 0.153 0.748 23.800 6.400
NaiveBayes 0.488 0.869 0.841 0.899 0.000 0.000 0.000 -0.127 0.768 27.600 2.600
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
MLP 7 15 9 120
KNN 3 19 7 122
RandomForest 5 17 9 120
SVM 13 9 64 65
DecisionTree 7 15 17 112
LogisticRegression 14 8 71 58
XGBoost 8 14 24 105
NaiveBayes 0 22 13 116
[LOSO] skipped (--skip-loso)
WCST_PCT_CONCEPTUAL_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_2 --allow-cols BAI_T1,BDI_T1,WCST_PCT_CONCEPTUAL_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'WCST_PCT_CONCEPTUAL_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_2
0 22 14.6
1 129 85.4
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.599 0.886 0.867 0.907 0.211 0.250 0.182 0.102 0.801 27.000 3.200
KNN 0.528 0.924 0.865 0.992 0.160 0.667 0.091 0.210 0.861 29.600 0.600
DecisionTree 0.487 0.867 0.851 0.884 0.103 0.118 0.091 -0.028 0.768 26.800 3.400
RandomForest 0.482 0.905 0.855 0.961 0.071 0.167 0.045 0.012 0.828 29.000 1.200
XGBoost 0.469 0.802 0.839 0.767 0.109 0.091 0.136 -0.082 0.675 23.600 6.600
LogisticRegression 0.429 0.709 0.840 0.612 0.177 0.123 0.318 -0.051 0.570 18.800 11.400
SVM 0.409 0.723 0.853 0.628 0.205 0.143 0.364 -0.006 0.589 19.000 11.200
NaiveBayes 0.297 0.894 0.847 0.946 0.000 0.000 0.000 -0.091 0.808 28.800 1.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
MLP 4 18 12 117
KNN 2 20 1 128
DecisionTree 2 20 15 114
RandomForest 1 21 5 124
XGBoost 3 19 30 99
LogisticRegression 7 15 50 79
SVM 8 14 48 81
NaiveBayes 0 22 7 122
[LOSO] skipped (--skip-loso)
WCST_TOTAL_ERRORS_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_2 --allow-cols BAI_T1,BDI_T1,WCST_TOTAL_ERRORS_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'WCST_TOTAL_ERRORS_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_2
0 22 14.6
1 129 85.4
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.530 0.920 0.864 0.984 0.154 0.500 0.091 0.166 0.854 29.400 0.800
DecisionTree 0.503 0.884 0.855 0.915 0.114 0.154 0.091 0.007 0.795 27.600 2.600
RandomForest 0.487 0.914 0.852 0.984 0.000 0.000 0.000 -0.048 0.841 29.800 0.400
XGBoost 0.461 0.808 0.874 0.752 0.258 0.200 0.364 0.092 0.695 22.200 8.000
MLP 0.452 0.875 0.853 0.899 0.108 0.133 0.091 -0.012 0.781 27.200 3.000
LogisticRegression 0.423 0.700 0.830 0.605 0.152 0.105 0.273 -0.089 0.556 18.800 11.400
SVM 0.384 0.730 0.871 0.628 0.250 0.172 0.455 0.060 0.603 18.600 11.600
NaiveBayes 0.322 0.902 0.849 0.961 0.000 0.000 0.000 -0.076 0.821 29.200 1.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
KNN 2 20 2 127
DecisionTree 2 20 11 118
RandomForest 0 22 2 127
XGBoost 8 14 32 97
MLP 2 20 13 116
LogisticRegression 6 16 51 78
SVM 10 12 48 81
NaiveBayes 0 22 5 124
[LOSO] skipped (--skip-loso)
cdrisc_lcga_group_3
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_3 --allow-cols BAI_T1,BDI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_3
0 54 35.8
1 59 39.1
2 38 25.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.575 0.427 0.432 0.424 0.406 0.409 0.404 0.126 0.424 11.000 11.000
DecisionTree 0.562 0.442 0.442 0.444 0.423 0.423 0.424 0.151 0.444 11.200 11.800
RandomForest 0.541 0.408 0.407 0.411 0.394 0.395 0.395 0.097 0.411 11.600 12.000
SVM 0.506 0.315 0.314 0.318 0.301 0.301 0.301 -0.047 0.318 12.200 11.600
MLP 0.493 0.391 0.393 0.391 0.382 0.382 0.384 0.078 0.391 10.200 11.600
KNN 0.446 0.305 0.290 0.338 0.277 0.268 0.305 -0.043 0.338 13.000 15.000
NaiveBayes 0.443 0.360 0.367 0.371 0.343 0.360 0.348 0.021 0.371 12.200 14.200
LogisticRegression 0.433 0.314 0.329 0.318 0.306 0.316 0.312 -0.030 0.318 7.600 14.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 22 14 18
True_1 14 32 13
True_2 19 9 10
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 28 13 13
True_1 17 29 13
True_2 14 14 10
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 25 17 12
True_1 21 27 11
True_2 14 14 10
[SVM]
Pred_0 Pred_1 Pred_2
True_0 16 24 14
True_1 23 25 11
True_2 19 12 7
[MLP]
Pred_0 Pred_1 Pred_2
True_0 26 17 11
True_1 19 21 19
True_2 13 13 12
[KNN]
Pred_0 Pred_1 Pred_2
True_0 27 24 3
True_1 29 23 7
True_2 19 18 1
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 24 23 7
True_1 27 26 6
True_2 20 12 6
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 24 14 16
True_1 28 15 16
True_2 20 9 9
[LOSO] skipped (--skip-loso)
EF_ENV_MONITOR
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_3 --allow-cols BAI_T1,BDI_T1,EF_ENV_MONITOR --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'EF_ENV_MONITOR']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_3
0 54 35.8
1 59 39.1
2 38 25.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
RandomForest 0.551 0.432 0.429 0.437 0.414 0.413 0.416 0.136 0.437 12.400 11.400
DecisionTree 0.539 0.394 0.391 0.397 0.380 0.380 0.382 0.076 0.397 12.000 11.600
XGBoost 0.522 0.355 0.352 0.358 0.333 0.331 0.334 0.017 0.358 12.400 10.800
LogisticRegression 0.507 0.371 0.374 0.371 0.357 0.357 0.359 0.048 0.371 10.000 11.800
NaiveBayes 0.498 0.375 0.380 0.391 0.358 0.373 0.367 0.054 0.391 11.800 14.600
MLP 0.496 0.339 0.343 0.338 0.325 0.326 0.325 -0.003 0.338 10.400 11.400
KNN 0.489 0.319 0.302 0.351 0.290 0.276 0.319 -0.013 0.351 11.400 15.600
SVM 0.448 0.354 0.378 0.358 0.340 0.358 0.348 0.037 0.358 6.800 14.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 28 15 11
True_1 18 29 12
True_2 11 18 9
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 28 17 9
True_1 21 23 15
True_2 9 20 9
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 20 16 18
True_1 20 28 11
True_2 14 18 6
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 27 14 13
True_1 19 20 20
True_2 13 16 9
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 29 19 6
True_1 28 24 7
True_2 16 16 6
[MLP]
Pred_0 Pred_1 Pred_2
True_0 23 15 16
True_1 21 20 18
True_2 13 17 8
[KNN]
Pred_0 Pred_1 Pred_2
True_0 31 17 6
True_1 29 21 9
True_2 18 19 1
[SVM]
Pred_0 Pred_1 Pred_2
True_0 29 7 18
True_1 25 16 18
True_2 18 11 9
[LOSO] skipped (--skip-loso)
HRV_RMSSD_MS
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_3 --allow-cols BAI_T1,BDI_T1,HRV_RMSSD_MS --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'HRV_RMSSD_MS']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_3
0 54 35.8
1 59 39.1
2 38 25.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.594 0.450 0.450 0.450 0.444 0.445 0.443 0.160 0.450 12.000 11.000
KNN 0.547 0.395 0.389 0.417 0.367 0.370 0.384 0.093 0.417 12.800 13.800
RandomForest 0.543 0.380 0.378 0.391 0.363 0.366 0.368 0.058 0.391 15.000 9.400
DecisionTree 0.524 0.375 0.374 0.377 0.363 0.365 0.363 0.044 0.377 12.600 11.200
XGBoost 0.511 0.349 0.348 0.351 0.336 0.336 0.337 0.008 0.351 12.400 10.600
SVM 0.488 0.297 0.302 0.305 0.289 0.291 0.298 -0.052 0.305 8.200 14.000
NaiveBayes 0.455 0.328 0.336 0.344 0.313 0.329 0.324 -0.020 0.344 11.000 15.600
LogisticRegression 0.439 0.330 0.334 0.331 0.322 0.324 0.326 -0.012 0.331 9.200 12.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[MLP]
Pred_0 Pred_1 Pred_2
True_0 23 20 11
True_1 19 30 10
True_2 13 10 15
[KNN]
Pred_0 Pred_1 Pred_2
True_0 29 17 8
True_1 23 30 6
True_2 17 17 4
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 18 23 13
True_1 18 33 8
True_2 11 19 8
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 19 23 12
True_1 21 28 10
True_2 16 12 10
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 18 21 15
True_1 21 26 12
True_2 14 15 9
[SVM]
Pred_0 Pred_1 Pred_2
True_0 25 15 14
True_1 28 13 18
True_2 17 13 8
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 27 20 7
True_1 33 20 6
True_2 18 15 5
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 23 15 16
True_1 27 17 15
True_2 14 14 10
[LOSO] skipped (--skip-loso)
WCST_NONPERS_ERR_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_3 --allow-cols BAI_T1,BDI_T1,WCST_NONPERS_ERR_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'WCST_NONPERS_ERR_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_3
0 54 35.8
1 59 39.1
2 38 25.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
SVM 0.549 0.241 0.244 0.238 0.230 0.231 0.228 -0.155 0.238 10.800 11.200
XGBoost 0.540 0.384 0.384 0.384 0.367 0.368 0.367 0.060 0.384 11.400 11.600
RandomForest 0.507 0.363 0.354 0.377 0.336 0.333 0.347 0.035 0.377 13.400 12.000
DecisionTree 0.503 0.344 0.344 0.344 0.327 0.327 0.328 0.001 0.344 12.200 10.200
KNN 0.461 0.341 0.340 0.371 0.313 0.320 0.338 0.015 0.371 11.400 16.400
MLP 0.459 0.314 0.316 0.311 0.296 0.298 0.295 -0.046 0.311 11.800 10.200
LogisticRegression 0.406 0.274 0.288 0.285 0.268 0.276 0.281 -0.075 0.285 6.400 14.800
NaiveBayes 0.382 0.291 0.292 0.305 0.275 0.283 0.284 -0.085 0.305 12.200 14.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[SVM]
Pred_0 Pred_1 Pred_2
True_0 12 24 18
True_1 24 18 17
True_2 20 12 6
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 21 16 17
True_1 21 28 10
True_2 16 13 9
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 23 19 12
True_1 21 30 8
True_2 16 18 4
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 17 17 20
True_1 21 27 11
True_2 13 17 8
[KNN]
Pred_0 Pred_1 Pred_2
True_0 29 19 6
True_1 30 25 4
True_2 23 13 2
[MLP]
Pred_0 Pred_1 Pred_2
True_0 20 17 17
True_1 20 21 18
True_2 11 21 6
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 25 11 18
True_1 30 10 19
True_2 19 11 8
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 22 24 8
True_1 32 20 7
True_2 17 17 4
[LOSO] skipped (--skip-loso)
WCST_PCT_PERS_ERR_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_3 --allow-cols BAI_T1,BDI_T1,WCST_PCT_PERS_ERR_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'WCST_PCT_PERS_ERR_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_3
0 54 35.8
1 59 39.1
2 38 25.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
RandomForest 0.552 0.415 0.415 0.424 0.397 0.405 0.400 0.107 0.424 14.600 10.600
XGBoost 0.544 0.406 0.403 0.411 0.386 0.385 0.388 0.095 0.411 12.800 10.800
SVM 0.533 0.251 0.260 0.272 0.237 0.245 0.256 -0.109 0.272 15.800 4.400
DecisionTree 0.526 0.371 0.372 0.371 0.353 0.354 0.353 0.042 0.371 12.200 10.200
MLP 0.501 0.362 0.360 0.364 0.349 0.350 0.349 0.025 0.364 13.000 10.400
NaiveBayes 0.471 0.317 0.332 0.358 0.295 0.326 0.323 -0.022 0.358 20.200 7.200
KNN 0.453 0.282 0.270 0.311 0.256 0.250 0.281 -0.087 0.311 13.000 15.000
LogisticRegression 0.431 0.291 0.304 0.298 0.289 0.295 0.301 -0.050 0.298 6.800 13.600
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 25 22 7
True_1 18 31 10
True_2 10 20 8
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 24 19 11
True_1 15 30 14
True_2 15 15 8
[SVM]
Pred_0 Pred_1 Pred_2
True_0 5 28 21
True_1 8 29 22
True_2 9 22 7
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 22 17 15
True_1 17 26 16
True_2 12 18 8
[MLP]
Pred_0 Pred_1 Pred_2
True_0 20 23 11
True_1 19 26 14
True_2 13 16 9
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 11 37 6
True_1 16 39 4
True_2 9 25 4
[KNN]
Pred_0 Pred_1 Pred_2
True_0 24 27 3
True_1 30 22 7
True_2 21 16 1
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 23 13 18
True_1 28 11 20
True_2 17 10 11
[LOSO] skipped (--skip-loso)
IGT_NET_TOTAL
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target cdrisc_lcga_group_3 --allow-cols BAI_T1,BDI_T1,IGT_NET_TOTAL --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=cdrisc_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'IGT_NET_TOTAL']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
cdrisc_lcga_group_3
0 54 35.8
1 59 39.1
2 38 25.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.536 0.377 0.376 0.377 0.355 0.355 0.355 0.050 0.377 11.800 11.000
KNN 0.530 0.375 0.381 0.404 0.347 0.368 0.369 0.065 0.404 13.400 14.600
DecisionTree 0.508 0.346 0.337 0.358 0.317 0.311 0.326 0.009 0.358 12.800 11.800
SVM 0.490 0.339 0.353 0.344 0.324 0.335 0.330 0.004 0.344 8.000 15.200
RandomForest 0.481 0.347 0.343 0.358 0.322 0.322 0.330 0.006 0.358 11.600 13.800
NaiveBayes 0.470 0.367 0.373 0.384 0.346 0.362 0.358 0.044 0.384 11.000 15.600
LogisticRegression 0.446 0.346 0.350 0.351 0.327 0.330 0.332 0.010 0.351 9.600 13.800
MLP 0.402 0.226 0.226 0.225 0.215 0.215 0.215 -0.181 0.225 11.000 11.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 22 17 15
True_1 16 28 15
True_2 17 14 7
[KNN]
Pred_0 Pred_1 Pred_2
True_0 28 21 5
True_1 26 30 3
True_2 19 16 3
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 22 17 15
True_1 20 29 10
True_2 17 18 3
[SVM]
Pred_0 Pred_1 Pred_2
True_0 27 13 14
True_1 27 18 14
True_2 22 9 7
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 23 17 14
True_1 26 27 6
True_2 20 14 4
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 29 17 8
True_1 30 24 5
True_2 19 14 5
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 26 15 13
True_1 23 21 15
True_2 20 12 6
[MLP]
Pred_0 Pred_1 Pred_2
True_0 13 25 16
True_1 27 16 16
True_2 19 14 5
[LOSO] skipped (--skip-loso)
isi_lcga_group_2
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_2 --allow-cols BAI_T1,BDI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_2 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_2
0 39 25.8
1 112 74.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.601 0.732 0.772 0.696 0.360 0.320 0.410 0.099 0.623 20.200 10.000
KNN 0.536 0.847 0.755 0.964 0.170 0.500 0.103 0.131 0.742 28.600 1.600
LogisticRegression 0.535 0.712 0.771 0.661 0.362 0.309 0.436 0.088 0.603 19.200 11.000
RandomForest 0.535 0.786 0.754 0.821 0.265 0.310 0.231 0.058 0.669 24.400 5.800
DecisionTree 0.534 0.750 0.750 0.750 0.282 0.282 0.282 0.032 0.629 22.400 7.800
NaiveBayes 0.510 0.825 0.731 0.946 0.000 0.000 0.000 -0.120 0.702 29.000 1.200
SVM 0.498 0.757 0.794 0.723 0.409 0.367 0.462 0.173 0.656 20.400 9.800
MLP 0.484 0.781 0.767 0.795 0.324 0.343 0.308 0.106 0.669 23.200 7.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
XGBoost 16 23 34 78
KNN 4 35 4 108
LogisticRegression 17 22 38 74
RandomForest 9 30 20 92
DecisionTree 11 28 28 84
NaiveBayes 0 39 6 106
SVM 18 21 31 81
MLP 12 27 23 89
[LOSO] skipped (--skip-loso)
IGT_NET_1
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_2 --allow-cols BAI_T1,BDI_T1,IGT_NET_1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'IGT_NET_1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_2
0 39 25.8
1 112 74.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.640 0.826 0.756 0.911 0.218 0.375 0.154 0.092 0.715 27.000 3.200
RandomForest 0.591 0.802 0.746 0.866 0.200 0.286 0.154 0.025 0.682 26.000 4.200
MLP 0.572 0.788 0.781 0.795 0.368 0.378 0.359 0.156 0.682 22.800 7.400
XGBoost 0.567 0.712 0.771 0.661 0.362 0.309 0.436 0.088 0.603 19.200 11.000
LogisticRegression 0.557 0.673 0.786 0.589 0.396 0.313 0.538 0.113 0.576 16.800 13.400
DecisionTree 0.539 0.764 0.761 0.768 0.312 0.316 0.308 0.076 0.649 22.600 7.600
NaiveBayes 0.521 0.813 0.734 0.911 0.078 0.167 0.051 -0.061 0.689 27.800 2.400
SVM 0.458 0.729 0.813 0.661 0.444 0.367 0.564 0.201 0.636 18.200 12.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
KNN 6 33 10 102
RandomForest 6 33 15 97
MLP 14 25 23 89
XGBoost 17 22 38 74
LogisticRegression 21 18 46 66
DecisionTree 12 27 26 86
NaiveBayes 2 37 10 102
SVM 22 17 38 74
[LOSO] skipped (--skip-loso)
DBAS
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_2 --allow-cols BAI_T1,BDI_T1,DBAS --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'DBAS']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_2
0 39 25.8
1 112 74.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.580 0.835 0.777 0.902 0.333 0.476 0.256 0.200 0.735 26.000 4.200
DecisionTree 0.572 0.782 0.779 0.786 0.364 0.368 0.359 0.146 0.675 22.600 7.600
RandomForest 0.566 0.831 0.771 0.902 0.305 0.450 0.231 0.171 0.728 26.200 4.000
XGBoost 0.559 0.752 0.774 0.732 0.357 0.333 0.385 0.112 0.642 21.200 9.000
NaiveBayes 0.555 0.829 0.748 0.929 0.157 0.333 0.103 0.050 0.715 27.800 2.400
LogisticRegression 0.534 0.643 0.750 0.562 0.340 0.269 0.462 0.021 0.536 16.800 13.400
MLP 0.501 0.781 0.767 0.795 0.324 0.343 0.308 0.106 0.669 23.200 7.000
SVM 0.450 0.696 0.758 0.643 0.337 0.286 0.410 0.048 0.583 19.000 11.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
KNN 10 29 11 101
DecisionTree 14 25 24 88
RandomForest 9 30 11 101
XGBoost 15 24 30 82
NaiveBayes 4 35 8 104
LogisticRegression 18 21 49 63
MLP 12 27 23 89
SVM 16 23 40 72
[LOSO] skipped (--skip-loso)
IGT_NET_TOTAL
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_2 --allow-cols BAI_T1,BDI_T1,IGT_NET_TOTAL --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'IGT_NET_TOTAL']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_2
0 39 25.8
1 112 74.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.583 0.759 0.759 0.759 0.308 0.308 0.308 0.067 0.642 22.400 7.800
LogisticRegression 0.560 0.702 0.774 0.643 0.371 0.310 0.462 0.094 0.596 18.600 11.600
RandomForest 0.537 0.806 0.735 0.893 0.111 0.200 0.077 -0.044 0.682 27.200 3.000
KNN 0.514 0.815 0.743 0.902 0.148 0.267 0.103 0.006 0.695 27.200 3.000
DecisionTree 0.509 0.778 0.746 0.812 0.235 0.276 0.205 0.020 0.656 24.400 5.800
SVM 0.494 0.682 0.727 0.643 0.264 0.231 0.308 -0.046 0.556 19.800 10.400
NaiveBayes 0.456 0.830 0.745 0.938 0.122 0.300 0.077 0.025 0.715 28.200 2.000
MLP 0.408 0.727 0.706 0.750 0.113 0.125 0.103 -0.158 0.583 23.800 6.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
XGBoost 12 27 27 85
LogisticRegression 18 21 40 72
RandomForest 3 36 12 100
KNN 4 35 11 101
DecisionTree 8 31 21 91
SVM 12 27 40 72
NaiveBayes 3 36 7 105
MLP 4 35 28 84
[LOSO] skipped (--skip-loso)
HRV_NN50
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_2 --allow-cols BAI_T1,BDI_T1,HRV_NN50 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'HRV_NN50']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_2
0 39 25.8
1 112 74.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.557 0.737 0.762 0.714 0.329 0.304 0.359 0.070 0.623 21.000 9.200
KNN 0.547 0.782 0.725 0.848 0.102 0.150 0.077 -0.097 0.649 26.200 4.000
RandomForest 0.541 0.788 0.736 0.848 0.164 0.227 0.128 -0.029 0.662 25.800 4.400
DecisionTree 0.532 0.731 0.760 0.705 0.326 0.298 0.359 0.061 0.616 20.800 9.400
LogisticRegression 0.531 0.700 0.795 0.625 0.412 0.333 0.538 0.145 0.603 17.600 12.600
MLP 0.522 0.751 0.761 0.741 0.321 0.310 0.333 0.073 0.636 21.800 8.400
NaiveBayes 0.514 0.825 0.731 0.946 0.000 0.000 0.000 -0.120 0.702 29.000 1.200
SVM 0.470 0.735 0.815 0.670 0.449 0.373 0.564 0.210 0.642 18.400 11.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
XGBoost 14 25 32 80
KNN 3 36 17 95
RandomForest 5 34 17 95
DecisionTree 14 25 33 79
LogisticRegression 21 18 42 70
MLP 13 26 29 83
NaiveBayes 0 39 6 106
SVM 22 17 37 75
[LOSO] skipped (--skip-loso)
WCST_TOTAL_ERRORS_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_2 --allow-cols BAI_T1,BDI_T1,WCST_TOTAL_ERRORS_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_2 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'WCST_TOTAL_ERRORS_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_2
0 39 25.8
1 112 74.2
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.592 0.731 0.760 0.705 0.326 0.298 0.359 0.061 0.616 20.800 9.400
RandomForest 0.577 0.813 0.746 0.893 0.179 0.294 0.128 0.029 0.695 26.800 3.400
DecisionTree 0.530 0.779 0.756 0.804 0.282 0.312 0.256 0.064 0.662 23.800 6.400
KNN 0.509 0.810 0.741 0.893 0.145 0.250 0.103 -0.007 0.689 27.000 3.200
LogisticRegression 0.454 0.673 0.756 0.607 0.340 0.279 0.436 0.038 0.563 18.000 12.200
NaiveBayes 0.447 0.811 0.725 0.920 0.000 0.000 0.000 -0.149 0.682 28.400 1.800
SVM 0.428 0.699 0.766 0.643 0.354 0.298 0.436 0.071 0.589 18.800 11.400
MLP 0.389 0.759 0.720 0.804 0.123 0.154 0.103 -0.109 0.623 25.000 5.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
model TN_sum FP_sum FN_sum TP_sum
XGBoost 14 25 33 79
RandomForest 5 34 12 100
DecisionTree 10 29 22 90
KNN 4 35 12 100
LogisticRegression 17 22 44 68
NaiveBayes 0 39 9 103
SVM 17 22 40 72
MLP 4 35 22 90
[LOSO] skipped (--skip-loso)
isi_lcga_group_3
BASELINE
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_3 --allow-cols BAI_T1,BDI_T1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_3 列數=151 特徵數=2
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_3
0 37 24.5
1 30 19.9
2 84 55.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.547 0.459 0.451 0.477 0.375 0.385 0.380 0.074 0.477 3.200 8.000
KNN 0.545 0.457 0.453 0.464 0.395 0.397 0.395 0.073 0.464 6.000 6.200
XGBoost 0.538 0.449 0.440 0.464 0.369 0.375 0.369 0.050 0.464 4.600 6.400
RandomForest 0.537 0.445 0.440 0.450 0.363 0.365 0.363 0.053 0.450 5.800 6.600
DecisionTree 0.524 0.404 0.422 0.391 0.325 0.330 0.324 0.013 0.391 8.200 7.400
NaiveBayes 0.519 0.393 0.334 0.523 0.244 0.223 0.319 -0.068 0.523 0.000 1.600
LogisticRegression 0.505 0.322 0.403 0.318 0.315 0.341 0.358 0.018 0.318 13.400 8.800
SVM 0.467 0.354 0.443 0.344 0.349 0.382 0.387 0.063 0.344 14.200 8.000
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[MLP]
Pred_0 Pred_1 Pred_2
True_0 13 2 22
True_1 8 4 18
True_2 19 10 55
[KNN]
Pred_0 Pred_1 Pred_2
True_0 9 3 25
True_1 6 10 14
True_2 16 17 51
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 11 3 23
True_1 6 5 19
True_2 15 15 54
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 10 6 21
True_1 8 6 16
True_2 15 17 52
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 8 12 17
True_1 11 7 12
True_2 18 22 44
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 1 0 36
True_1 1 0 29
True_2 6 0 78
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 12 13 12
True_1 8 15 7
True_2 24 39 21
[SVM]
Pred_0 Pred_1 Pred_2
True_0 16 13 8
True_1 6 14 10
True_2 18 44 22
[LOSO] skipped (--skip-loso)
IGT_NET_1
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_3 --allow-cols BAI_T1,BDI_T1,IGT_NET_1 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'IGT_NET_1']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_3
0 37 24.5
1 30 19.9
2 84 55.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
DecisionTree 0.566 0.476 0.485 0.470 0.408 0.407 0.412 0.129 0.470 6.200 8.800
LogisticRegression 0.544 0.389 0.468 0.377 0.369 0.391 0.403 0.093 0.377 10.600 10.800
XGBoost 0.536 0.452 0.440 0.470 0.363 0.367 0.366 0.056 0.470 3.800 7.000
MLP 0.536 0.437 0.432 0.444 0.342 0.340 0.348 0.048 0.444 4.400 8.600
NaiveBayes 0.534 0.425 0.389 0.523 0.290 0.295 0.339 0.007 0.523 0.400 3.000
KNN 0.530 0.436 0.437 0.437 0.356 0.357 0.358 0.048 0.437 4.800 8.800
RandomForest 0.523 0.439 0.424 0.464 0.344 0.348 0.350 0.029 0.464 3.600 6.400
SVM 0.471 0.411 0.483 0.391 0.375 0.399 0.398 0.098 0.391 12.000 8.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 16 8 13
True_1 8 7 15
True_2 20 16 48
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 18 11 8
True_1 9 12 9
True_2 27 30 27
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 11 5 21
True_1 6 4 20
True_2 18 10 56
[MLP]
Pred_0 Pred_1 Pred_2
True_0 12 8 17
True_1 10 3 17
True_2 21 11 52
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 5 1 31
True_1 1 0 29
True_2 9 1 74
[KNN]
Pred_0 Pred_1 Pred_2
True_0 12 8 17
True_1 8 5 17
True_2 24 11 49
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 10 4 23
True_1 6 3 21
True_2 16 11 57
[SVM]
Pred_0 Pred_1 Pred_2
True_0 14 15 8
True_1 7 13 10
True_2 20 32 32
[LOSO] skipped (--skip-loso)
IGT_NET_5
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_3 --allow-cols BAI_T1,BDI_T1,IGT_NET_5 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'IGT_NET_5']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_3
0 37 24.5
1 30 19.9
2 84 55.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.551 0.485 0.478 0.497 0.406 0.409 0.409 0.124 0.497 4.200 7.800
XGBoost 0.545 0.488 0.483 0.497 0.415 0.420 0.413 0.123 0.497 5.200 6.600
RandomForest 0.541 0.474 0.463 0.497 0.386 0.397 0.387 0.088 0.497 4.000 5.800
SVM 0.540 0.375 0.448 0.371 0.371 0.389 0.415 0.090 0.371 12.200 9.400
LogisticRegression 0.524 0.322 0.419 0.338 0.339 0.365 0.406 0.076 0.338 11.600 12.400
MLP 0.515 0.439 0.442 0.437 0.362 0.362 0.362 0.054 0.437 6.000 7.800
NaiveBayes 0.498 0.398 0.345 0.510 0.253 0.238 0.316 -0.053 0.510 0.400 2.400
DecisionTree 0.489 0.384 0.394 0.377 0.305 0.310 0.304 -0.030 0.377 8.200 6.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[KNN]
Pred_0 Pred_1 Pred_2
True_0 15 3 19
True_1 8 5 17
True_2 16 13 55
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 13 4 20
True_1 6 7 17
True_2 14 15 55
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 12 4 21
True_1 4 4 22
True_2 13 12 59
[SVM]
Pred_0 Pred_1 Pred_2
True_0 17 10 10
True_1 6 15 9
True_2 24 36 24
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 22 10 5
True_1 7 13 10
True_2 33 35 16
[MLP]
Pred_0 Pred_1 Pred_2
True_0 10 7 20
True_1 10 7 13
True_2 19 16 49
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 2 0 35
True_1 3 0 27
True_2 7 2 75
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 7 10 20
True_1 9 6 15
True_2 15 25 44
[LOSO] skipped (--skip-loso)
CPT_OMISSION_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_3 --allow-cols BAI_T1,BDI_T1,CPT_OMISSION_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'CPT_OMISSION_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_3
0 37 24.5
1 30 19.9
2 84 55.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
KNN 0.543 0.464 0.463 0.470 0.394 0.403 0.392 0.084 0.470 4.200 8.200
XGBoost 0.541 0.457 0.448 0.470 0.364 0.365 0.367 0.075 0.470 5.800 5.800
RandomForest 0.513 0.421 0.404 0.470 0.313 0.329 0.328 -0.014 0.470 3.400 3.800
LogisticRegression 0.513 0.356 0.407 0.344 0.330 0.344 0.355 0.013 0.344 11.800 7.800
NaiveBayes 0.487 0.414 0.372 0.523 0.271 0.270 0.329 -0.004 0.523 0.600 2.400
MLP 0.482 0.402 0.394 0.411 0.315 0.313 0.318 -0.022 0.411 4.800 7.400
DecisionTree 0.474 0.371 0.366 0.377 0.278 0.278 0.280 -0.071 0.377 6.600 5.800
SVM 0.438 0.395 0.437 0.377 0.340 0.354 0.350 0.035 0.377 11.000 6.800
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[KNN]
Pred_0 Pred_1 Pred_2
True_0 12 4 21
True_1 7 7 16
True_2 22 10 52
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 7 9 21
True_1 8 7 15
True_2 14 13 57
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 5 3 29
True_1 4 3 23
True_2 10 11 63
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 11 10 16
True_1 8 13 9
True_2 20 36 28
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 3 2 32
True_1 2 0 28
True_2 7 1 76
[MLP]
Pred_0 Pred_1 Pred_2
True_0 10 6 21
True_1 7 3 20
True_2 20 15 49
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 5 8 24
True_1 9 4 17
True_2 15 21 48
[SVM]
Pred_0 Pred_1 Pred_2
True_0 9 15 13
True_1 7 11 12
True_2 18 29 37
[LOSO] skipped (--skip-loso)
PSQI_T2
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_3 --allow-cols BAI_T1,BDI_T1,PSQI_T2 --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'PSQI_T2']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_3
0 37 24.5
1 30 19.9
2 84 55.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
XGBoost 0.544 0.476 0.468 0.490 0.401 0.408 0.400 0.097 0.490 4.400 6.600
SVM 0.532 0.344 0.402 0.325 0.312 0.334 0.326 -0.011 0.325 12.200 7.800
MLP 0.510 0.450 0.443 0.457 0.367 0.368 0.367 0.060 0.457 5.200 7.000
KNN 0.506 0.423 0.410 0.444 0.332 0.335 0.338 0.003 0.444 3.400 7.200
LogisticRegression 0.502 0.369 0.459 0.364 0.364 0.391 0.411 0.093 0.364 12.400 10.000
RandomForest 0.490 0.459 0.439 0.503 0.348 0.358 0.361 0.068 0.503 2.800 5.200
DecisionTree 0.485 0.380 0.394 0.371 0.324 0.325 0.327 -0.022 0.371 7.200 8.400
NaiveBayes 0.454 0.399 0.347 0.503 0.260 0.243 0.317 -0.060 0.503 0.000 3.200
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 14 1 22
True_1 6 5 19
True_2 13 16 55
[SVM]
Pred_0 Pred_1 Pred_2
True_0 12 13 12
True_1 8 10 12
True_2 19 38 27
[MLP]
Pred_0 Pred_1 Pred_2
True_0 10 5 22
True_1 9 6 15
True_2 16 15 53
[KNN]
Pred_0 Pred_1 Pred_2
True_0 10 4 23
True_1 6 3 21
True_2 20 10 54
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 17 12 8
True_1 7 15 8
True_2 26 35 23
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 9 3 25
True_1 7 2 21
True_2 10 9 65
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 8 8 21
True_1 8 9 13
True_2 26 19 39
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 3 0 34
True_1 2 0 28
True_2 11 0 73
[LOSO] skipped (--skip-loso)
WCST_TOTAL_ERRORS_T
$ /Users/yuchi/PycharmProjects/PsyMl_ISI/.venv/bin/python '/Users/yuchi/PycharmProjects/PsyMl_ISI/學長/ml_benchmark_modular.py' --target isi_lcga_group_3 --allow-cols BAI_T1,BDI_T1,WCST_TOTAL_ERRORS_T --folds 5 --seed 42 --skip-loso
[資料] 來源=isi_raw_data 目標=isi_lcga_group_3 列數=151 特徵數=3
[特徵] 使用欄位（前 15）：['BDI_T1', 'BAI_T1', 'WCST_TOTAL_ERRORS_T']
[CV] Stratified 5-fold, seed=42 | class_weight=balanced
[聚合] K-fold = weighted | LOSO = OFF
[Leakage check] Class balance
count percent%
isi_lcga_group_3
0 37 24.5
1 30 19.9
2 84 55.6
[Leakage check] 未發現與目標 |r| ≥ 0.95 的欄位。
=== Basic ML Benchmark (Stratified 5-fold CV) ===
model AUC F1_pos(=1) Prec_pos Rec_pos F1_neg(=0) Prec_neg Rec_neg MCC Accuracy Pred1_mean Pred0_mean
MLP 0.585 0.494 0.487 0.503 0.397 0.397 0.398 0.140 0.503 5.200 7.000
KNN 0.575 0.470 0.448 0.497 0.360 0.349 0.375 0.106 0.497 3.200 7.800
SVM 0.575 0.357 0.426 0.338 0.323 0.349 0.344 0.018 0.338 12.800 7.600
XGBoost 0.563 0.449 0.434 0.470 0.348 0.349 0.354 0.054 0.470 4.200 6.400
RandomForest 0.561 0.449 0.433 0.483 0.346 0.358 0.354 0.045 0.483 3.400 5.400
NaiveBayes 0.531 0.434 0.398 0.550 0.291 0.301 0.350 0.062 0.550 0.000 2.400
DecisionTree 0.505 0.413 0.415 0.411 0.338 0.340 0.337 0.008 0.411 6.600 7.000
LogisticRegression 0.502 0.350 0.445 0.344 0.343 0.375 0.389 0.067 0.344 13.200 9.400
--- Aggregated Confusion Matrix Sums (across all folds' test parts) ---
多分類 Confusion Matrix（預測 vs 真實）：
[MLP]
Pred_0 Pred_1 Pred_2
True_0 12 8 17
True_1 11 5 14
True_2 12 13 59
[KNN]
Pred_0 Pred_1 Pred_2
True_0 14 3 20
True_1 13 1 16
True_2 12 12 60
[SVM]
Pred_0 Pred_1 Pred_2
True_0 11 15 11
True_1 8 12 10
True_2 19 37 28
[XGBoost]
Pred_0 Pred_1 Pred_2
True_0 10 4 23
True_1 10 3 17
True_2 12 14 58
[RandomForest]
Pred_0 Pred_1 Pred_2
True_0 7 2 28
True_1 9 4 17
True_2 11 11 62
[NaiveBayes]
Pred_0 Pred_1 Pred_2
True_0 4 0 33
True_1 3 0 27
True_2 5 0 79
[DecisionTree]
Pred_0 Pred_1 Pred_2
True_0 11 7 19
True_1 7 5 18
True_2 17 21 46
[LogisticRegression]
Pred_0 Pred_1 Pred_2
True_0 15 14 8
True_1 7 15 8
True_2 25 37 22
[LOSO] skipped (--skip-loso)