Development of transient ischemic attack risk prediction model suitable for initializing a learning health system unit using electronic medical records

Table 3 Comparison of different ML algorithms for TIA risk prediction models at various development stages

Development Stage	Performance Metrics ^*	ML Algorithm
Development Stage	Performance Metrics ^*	XGBoost	Random Forest	Support Vector Machines	K-Nearest Neighbors
Inclusive model: 150 variables	Recall	0.868	0.816	0.855	0.829
	Precision	0.815	0.879	0.833	0.621
	ROC-AUC	0.882	0.878	0.882	0.779
	Accuracy	0.886	0.897	0.890	0.764
Practical model: 20 variables	Recall	0.855	0.829	0.816	0.711
	Precision	0.660	0.663	0.660	0.610
	ROC-AUC	0.810	0.802	0.796	0.734
	Accuracy	0.796	0.794	0.789	0.741
Practical model in initializing ML-LHS: 20 variables	Recall	0.830
	Precision	0.726
	ROC-AUC	0.816
	Accuracy	0.812

ISSN: 1472-6947