Table 6. Results for different workflows of random forests for hospital readmissions data, with standard deviations over n = 10 runs.

RF Workflow	ROC AUC	Sensitivity	Specificity	Accuracy
No Transformation	0.60 (0.047)	0.58(0.042)	0.57 (0.049)	0.57 (0.046)
No Transformation, SMOTE	0.52 (0.053)	0.46 (0.052)	0.75 (0.055)	0.70 (0.055)
No Transformation, ROSE	0.5 (0)	0 (0)	1(0)	0.827 (0)
PCA	0.56(0.051)	0.50 (0.066)	0.63 (0.068)	0.61 (0.068)
PCA, SMOTE	0.57 (0.053)	0.62(0.051)	0.60 (0.058)	0.60(0.056)
PCA, ROSE	0.53 (0.072)	0.54 (0.071)	0.56 (0.076)	0.56 (0.075)
Mapper, No Transformations	0.49 (0.078)	0.46 (0.084)	0.60 (0.081)	0.58 (0.082)
Mapper, No Transformations, SMOTE	0.55 (0.087)	0.58 (0.075)	0.54 (0.082)	0.55 (0.080)
Mapper, No Transformation, ROSE	0.51 (0.093)	0.54(0.116)	0.51 (0.143)	0.52 (0.137)
Mapper, Node PCA	0.57 (0.069)	0.62 (0.076)	0.62 (0.086)	0.62 (0.083)
Mapper, Node PCA, SMOTE	0.57 (0.084)	0.46 (0.095)	0.71 (0.091)	0.67 (0.092)
Mapper, Node PCA, ROSE	0.64 (0.110)	0.65 (0.099)	0.61 (0.091)	0.62 (0.097)