Skip to main content
Fig. 5 | Journal of Translational Medicine

Fig. 5

From: Prediction of postpartum depression in women: development and validation of multiple machine learning models

Fig. 5

The feature importance and interpretation of the PP-PPD model (based on the ANN algorithm). A and E The importance ranking of features based on the mean (|SHAP value|). B and F A summary plot of the SHAP values for each feature. C and G SHAP force plot for a woman (PPD). D and H: SHAP force plot for a woman (without PPD). The higher the SHAP value of the feature, the impact of the feature on the model is larger. The red dots in the feature value represent higher values for that individual patient, whereas the blue dots indicate lower feature values. MCPP mother-in-law’s care in postpartum, BDI.1 Beck depression inventory(> 13 scores), BAI.1 Beck anxiety inventory (> 7 scores), EPQ-N.2 neuroticism dimension of the Eysenck personality questionnaire (> 56.7 scores), EPQ-P.2 psychoticism dimension of the Eysenck personality questionnaire (> 56.7 scores), PCPP.2 primary caregiver in postpartum (husband), MT.1 melancholic temperament (yes), PCPP.3 primary caregiver in postpartum (confinement nurse); Primopara.1, (yes); FT3 serum-free triiodothyronine, EPQ-N.1 neuroticism dimension of the Eysenck personality questionnaire (43.3–56.7 scores), EMSS.2 Enrich marital satisfaction scale (< 30 scores), PPPC.1 primary caregiver in postpartum (mother), EMSS.1 Enrich marital satisfaction scale (30–42 scores), EPQ-P.1 psychoticism dimension of the Eysenck personality questionnaire (43.3–56.7 scores), TC total cholesterol

Back to article page