Figure 3

Construction and evaluation of RF, SVM, and LASSO machine models. (A,B) Error graph of the random forest (RF) models, the genes are screened as meanDecreaseGini increases. (C) Support vector machine (SVM) models, the gene number of 4 is decided as RMSE (cross-validation) went the lowest. (D) Coefficients in the LASSO regression. (E) Cross-validation for tuning parameter selection in the proportional hazards model, the gene number of 9 is decided as Partial likelihood deviance went the lowest. (F) ROC analysis of LASSO models. (G,H) Survival analysis of different groups assessing the independence of the signatures. (I) Nomogram for predicting the proportion of patients. (J–L) The expression of prognosis-related DSRGs, risk score, and patients’ survival status between the high and low risk groups, the shade of the color illustrates the expression value of the genes. (M) VENN plot showing the intersection of genes screened by three machine models.