Figure 3

ML model synthesis analysis. (a) Training set ROC and AUC; (b) Test set ROC and AUC. Patients with KOA were sampled five times in an 8:2 ratio. (c) For the calibration curves of the test set, the horizontal coordinate is the average predicted probability, the vertical coordinate is the actual probability of the event, the diagonal dashed line is the reference line, and the other solid lines are the fitted lines of the different models. The closer the fitted line is to the reference line and the smaller the value in parentheses, the more accurate the model predictions are. (d) Test set DCA, where the black dashed line represents the assumption that all patients have severe osteoarthritis, and the red dashed line and the thin black line represent the assumption that patients do not have severe osteoarthritis. The remaining solid lines represent different models. (e, f) Training set, test set PR curves, and AP values. y-axis is precision and x-axis is recall.