Comparison of Machine Learning Methods and Conventional Logistic Regressions for Predicting Gestational Diabetes Using Routine Clinical Data: A Retrospective Cohort Study

Figure 3

Contribution of the predictor variables in GBDT and the logistic regression model. (a) Importance of the predictor variables in the GBDT model in the validation cohort. (b) Importance of the predictor variables in the logistic model with restricted cubic spline in the validation cohort. (c) Partial plot of the effects of fasting blood glucose (GLU, mmol/L), glycosylated hemoglobin (HbA1c, %), triglyceride (TG, mmol/L), and maternal BMI (kg/m²) on the risk of GDM across different values in the GBDT model. (d) Partial plot of the effect of glycosylated hemoglobin (HbA1c, %), high-density lipoprotein (HDL, mmol/L), fasting blood glucose (GLU, mmol/L), and triglyceride (TG, mmol/L) on the risk of GDM across different values in the logistic model with restricted cubic spline.

(a)

(b)

(c)

(d)