An Intelligent Diagnostic System to Analyze Early-Stage Chronic Kidney Disease for Clinical Application

Table 9

Experimental result analysis for clinical unseen dataset, i.e., training the model with CKD-15 and testing with CKD-21.

Classifier	Recall	TNR	PPV	F1-score	Train accuracy	Test accuracy	ROC-AUC
ADAB	0.98	0.95	0.95	0.96	100	96.48	0.965
LR	0.95	0.98	0.98	0.96	99.01	96.48	0.984
PAC	0.95	0.98	0.98	0.96	99.2	96.48	0.961
RF	0.97	0.95	0.95	0.96	100	96.09	0.961
SGD	0.97	0.95	0.95	0.96	99.8	96.09	0.938
MLP	0.96	0.95	0.95	0.95	100	95.7	0.957
SVM	0.97	0.94	0.94	0.95	99.2	95.31	0.953
NB	0.9	0.99	0.99	0.94	97.22	94.53	0.981
ELM	0.89	0.98	0.98	0.93	0.98	93.75	0.964
KNN	0.88	0.99	0.99	0.93	98.41	93.75	0.938
DT	0.97	0.89	0.9	0.93	100	92.97	0.93
GB	0.97	0.89	0.9	0.93	100	92.97	0.93

The best results are indicated in bold.