Table 2 Prehospital diagnostic algorithms for acute coronary syndrome using 43 features.

Models	AUC	Sensitivity	Accuracy	Specificity	F1-score	PPV	NPV
Training score
XGBoost	0.887	0.819	0.887	0.826	0.755	0.712	0.890
Logistic regression	0.893	0.818	0.893	0.807	0.762	0.698	0.906
Random forest	0.922	0.860	0.922	0.874	0.805	0.781	0.909
SVM (Linear)	0.894	0.822	0.894	0.823	0.762	0.713	0.898
SVM (radial basis function)	0.902	0.842	0.902	0.836	0.790	0.737	0.916
MLP	0.893	0.829	0.893	0.826	0.772	0.722	0.905
LDA	0.890	0.826	0.890	0.834	0.763	0.723	0.894
LGBM	0.894	0.823	0.894	0.819	0.764	0.711	0.902
Voting	0.927	0.852	0.927	0.839	0.804	0.744	0.928
Test score
XGBoost	0.849	0.756	0.792	0.811	0.715	0.684	0.864
Random forest	0.850	0.755	0.798	0.821	0.725	0.711	0.865
Logistic regression	0.843	0.740	0.780	0.801	0.703	0.693	0.857
SVM (Linear)	0.847	0.745	0.789	0.813	0.709	0.690	0.861
SVM (radial basis function)	0.834	0.735	0.791	0.821	0.708	0.687	0.855
MLP	0.834	0.709	0.786	0.826	0.695	0.695	0.846
LDA	0.860	0.761	0.802	0.823	0.727	0.706	0.870
LGBM	0.841	0.756	0.778	0.791	0.705	0.671	0.860
Voting	0.861	0.772	0.803	0.821	0.733	0.711	0.873
External cohort score
XGBoost	0.840	0.897	0.790	0.697	0.800	0.722	0.885
Random forest	0.803	0.690	0.726	0.758	0.702	0.714	0.735
Logistic regression	0.831	0.793	0.758	0.727	0.754	0.719	0.800
SVM (Linear)	0.838	0.793	0.758	0.727	0.754	0.719	0.800
SVM (radial basis function)	0.808	0.828	0.742	0.667	0.750	0.686	0.815
MLP	0.818	0.793	0.758	0.727	0.754	0.719	0.800
LDA	0.832	0.862	0.774	0.697	0.781	0.714	0.852
LGBM	0.789	0.552	0.742	0.909	0.667	0.842	0.698
Voting	0.828	0.862	0.790	0.727	0.794	0.735	0.857

AUC (area under the receiver operating characteristic curve), LDA (linear discriminant analysis), LGBM (light gradient boosting machine), MLP (multilayer perceptron), NPV (negative predictive values), PPV (posititive predictive values), SVM (support vector machine), XGBoost (eXtreme Gradient Boosting).

Quick links

Search