Design and Assessment of a Robust and Generalizable ANN-Based Classifier for the Prediction of Premature Birth by means of Multichannel Electrohysterographic Records

Table 4

Mean and standard deviation (between brackets) of different classifier metrics obtained for the test datasets for the 10 balanced data obtained using SMOTE techniques. For each balanced dataset, a cross validation method was carried out 30 times.

	Accuracy (%)	AUC (%)	Sensibility (%)	Specificity (%)	PPV (%)	NPV (%)
C1	73.2 (1.3)	76.5 (2.1)	70.5 (2.0)	75.3 (2.2)	74.8 (2.7)	73.6 (2.1)
C2	75.2 (1.5)	79.7 (2.3)	72.1 (2.1)	76.2 (2.4)	75.2 (2.7)	73.9 (2.2)
C3	80.2 (1.5)	83.7 (2.7)	76.7 (2.3)	82.6 (2.7)	79.8 (3.2)	77.6 (1.7)
C4	81.6 (1.5)	84.8 (2.8)	77.3 (1.9)	83.4 (2.5)	81.6 (3.3)	78.7 (1.8)
C5	82.8 (1.5)	84.7 (2.4)	79.3 (1.8)	84.1 (2.2)	83.6 (3.2)	80.1 (1.5)
C6	83.1 (1.6)	85.2 (2.6)	79.8 (1.9)	84.7 (2.0)	83.4 (2.9)	79.8 (2.2)
C7	81.7 (1.6)	86.6 (2.8)	79.7 (1.8)	85.2 (2.8)	84.5 (3.1)	80.1 (2.0)
C8	83.5 (1.6)	88.2 (2.9)	80.8 (1.8)	86.3 (3.1)	85.3 (2.7)	81.6 (1.9)
C9	85.8 (1.4)	88.7 (2.3)	81.8 (2.0)	87.3 (2.8)	86.3 (3.2)	81.2 (1.7)
C10	87.9 (1.6)	91.1 (2.5)	84.4 (2.1)	89.2 (2.6)	88.7 (2.7)	84.3 (1.9)