[Retracted] Constructing a Predictive Model of Depression in Chemotherapy Patients with Non-Hodgkin’s Lymphoma to Improve Medical Staffs’ Psychiatric Care

Figure 2

Prediction model selection. (a) Flow chart of model screening and evaluation design. (b) The support vector machine-recursive feature elimination (SVM-RFE) was used to build a prediction model. The SVM-RFE model had the highest accuracy () when it included 7 predictive factors. (c) Importance values of each factor in the random forest model (the picture above). And the ROC curve () demonstrated the accuracy of the random forest model (the picture below). (d) Based on the number of occurrences of each factor, it is incorporated into the logic model to obtain a pattern diagram of the AUC values (the picture above). Ultimately, the simplest predictive model with good predictive power can be constructed using only 10 predictors (the picture below).

(a)

(b)

(c)

(d)