Use of Machine Learning and Routine Laboratory Tests for Diabetes Mellitus Screening

Table 4

Performance of the studied regression models using healthy, prediabetes, and diabetes (HPDr) dataset.

Class	KNNr	SVMr	NBr	RFr	ANNr
Sensitivity (SN %)
Diabetes	51.3	67.0	63.1	68.0	66.3
Prediabetes	60.7	59.9	56.3	47.1	51.7
Healthy	75.2	80.2	79.4	85.8	85.3
Specificity (SP %)
Diabetes	98.7	98.0	95.9	97.9	98.0
Prediabetes	71.2	78.2	77.8	82.9	81.7
Healthy	71.9	73.1	72.6	63.5	67.5
Precision (PR %)
Diabetes	87.0	85.0	72.3	84.6	84.8
Prediabetes	56.0	62.4	60.6	62.5	63.1
Healthy	70.9	73.1	72.5	68.2	70.5
Negative precision (NPR %)
Diabetes	92.2	94.5	93.8	94.5	94.4
Prediabetes	74.9	76.3	74.6	72.1	73.6
Healthy	76.1	80.3	79.5	83.1	83.4
F1 score (F1 %)
Diabetes	64.5	74.9	67.4	74.8	73.4
Prediabetes	58.3	61.1	58.3	53.7	56.8
Healthy	73.0	76.5	75.8	73.0	77.2