Table 4 Performance metrics for nine ML models in the method 3 with eight selected features.

From: Machine learning models to predict osteoporosis in patients with chronic kidney disease stage 3–5 and end-stage kidney disease

Models

Datasets

Accuracy

Precision

Recall

F1 Score

AUC

LR

Validation

Value

0.765

0.787

0.900

0.840

0.796

95% CI

(0.739–0.790)

(0.758–0.814)

(0.879–0.921)

(0.820–0.860)

(0.766–0.825)

Test

Value

0.767

0.798

0.882

0.838

0.805

95% CI

(0.745–0.789)

(0.774–0.823)

(0.862–0.902)

(0.821–0.855)

(0.779–0.828)

XGBoost

Validation

Value

0.806

0.845

0.878

0.861

0.859

95% CI

(0.784–0.829)

(0.818–0.870)

(0.856–0.901)

(0.843–0.879)

(0.836–0.880)

Test

Value

0.823

0.854

0.894

0.873

0.878

95% CI

(0.803–0.843)

(0.830–0.876)

(0.874–0.914)

(0.857–0.889)

(0.858–0.897)

LightGBM

Validation

Value

0.802

0.838

0.881

0.859

0.862

95% CI

(0.777–0.825)

(0.811–0.864)

(0.858–0.903)

(0.839–0.878)

(0.837–0.884)

Test

Value

0.824

0.857

0.891

0.873

0.883

95% CI

(0.803–0.843)

(0.835–0.877)

(0.870–0.909)

(0.858–0.888)

(0.863–0.901)

CatBoost

Validation

Value

0.804

0.839

0.885

0.861

0.867

95% CI

(0.778–0.827)

(0.810–0.863)

(0.860–0.908)

(0.841–0.879)

(0.843–0.889)

Test

Value

0.838

0.864

0.905

0.884

0.888

95% CI

(0.818–0.858)

(0.841–0.886)

(0.886–0.925)

(0.869–0.900)

(0.869–0.906)

SVM

Validation

Value

0.781

0.806

0.894

0.848

0.811

95% CI

(0.755–0.806)

(0.779–0.836)

(0.872–0.914)

(0.829–0.868)

(0.781–0.840)

Test

Value

0.785

0.817

0.883

0.849

0.839

95% CI

(0.762–0.808)

(0.791–0.842)

(0.863–0.905)

(0.831–0.866)

(0.816–0.861)

DT

Validation

Value

0.717

0.801

0.781

0.791

0.679

95% CI

(0.689–0.743)

(0.771–0.832)

(0.752–0.811)

(0.767–0.815)

(0.650–0.710)

Test

Value

0.730

0.813

0.787

0.800

0.697

95% CI

(0.707–0.754)

(0.785–0.838)

(0.761–0.811)

(0.780–0.819)

(0.670–0.724)

RF

Validation

Value

0.798

0.827

0.891

0.858

0.850

95% CI

(0.772–0.822)

(0.799–0.854)

(0.867–0.913)

(0.838–0.877)

(0.825–0.874)

Test

Value

0.830

0.859

0.899

0.879

0.883

95% CI

(0.810–0.851)

(0.837–0.883)

(0.879–0.918)

(0.864–0.894)

(0.863–0.902)

KNN

Validation

Value

0.763

0.814

0.846

0.830

0.804

95% CI

(0.737–0.788)

(0.784–0.843)

(0.819–0.873)

(0.808–0.850)

(0.776–0.829)

Test

Value

0.772

0.822

0.849

0.836

0.823

95% CI

(0.751–0.795)

(0.799–0.847)

(0.827–0.872)

(0.818–0.854)

(0.801–0.845)

ANN

Validation

Value

0.873*

0.897*

0.920*

0.909*

0.940*

95% CI

(0.855–0.891)

(0.876–0.917)

(0.901–0.939)

(0.894–0.922)

(0.925–0.954)

Test

Value

0.861*

0.885*

0.916*

0.900*

0.930*

95% CI

(0.841–0.878)

(0.863–0.904)

(0.897–0.934)

(0.885–0.914)

(0.916–0.942)

  1. *The best result is indicated in bold. ANN, artificial neural network; AUC, area under the receiver operating characteristic curve; CatBoost, category boosting; CI, confidence interval; DT, decision tree; KNN, k-nearest neighbors algorithm; LightGBM, light gradient boosting machine; LR, logistic regression; RF, random forest; SVM, support vector machine; XGBoost, extreme gradient boosting.
Back to article page