Exploration of Black Boxes of Supervised Machine Learning Models: A Demonstration on Development of Predictive Heart Risk Score

Table 1

Performance assessment of ML models for the development of NLHRS.

Models	ANN		Linear SVM		RBF SVM		Random forest		LR
Confusion matrix	Case	Control	Case	Control	Case	Control	Case	Control	Case	Control
Case	185	45	187	43	187	43	191	39	188	42
Control	35	195	46	184	50	180	56	174	50	180
Accuracy	82.61		80.65		79.80		79.30		80.00
Sensitivity	0.791		0.813		0.813		0.830		0.817
Specificity	0.848		0.800		0.783		0.757		0.783
Kappa statistic	0.653		0.613		0.595		0.587		0.600
AUC	0.883		0.881		0.870		0.857		0.873
RMSE	0.365		0.372		0.379		0.411		0.380
Number of criteria fulfilled	5/6		5/6		1/6		1/6		Baseline risk prediction model