Skip to main content

Table 3 Evaluation of prediction performance of each model

From: Comparison of the effectiveness of different machine learning algorithms in predicting new fractures after PKP for osteoporotic vertebral compression fractures

Models

AUC (95% CI)

Kappa

Sensitivity

Specificity

Logistic regression

0.898 (0.860–0.936)

0.794

0.831

0.965

Random forest

0.940 (0.910–0.970)

0.880

0.966

0.913

Gradient boosting machine

0.910 (0.873–0.947)

0.820

0.898

0.922

Decision tree

0.842 (0.796–0.888)

0.683

0.779

0.904

Support vector machine

0.902 (0.864–0.940)

0.803

0.856

0.948

Neural network

0.923 (0.889–0.957)

0.846

0.907

0.939

Regularized discriminant analysis

0.915 (0.879–0.950)

0.828

0.881

0.947

  1. The Kappa values are based on the confusion matrix used for consistency testing, with results in the range of 0–0.2 for SLIGHT, 0.21–0.4 for FAIR, 0.41–0.60 for MODERATE, 0.61–0.80 for SUBSTANTIAL, and 0.81–1 for ALMOST PERFECT. Sensitivity is the percentage of true positive samples among actual positive samples, and specificity is the percentage of true negative samples among actual negative samples