Table 2.

Average Performance Metrics (Precision, Recall, MCC, Cohen’s Kappa, Sensitivity, Specificity, and AUC) for Each Machine Learning Modeling Predicting Surgery Ending by 5 pm and Patient Discharged From PACU by 7 pm Based on 1 pm, 2 pm, 3 pm, or 4 pm Start

1 pm start: predicting if surgery will end before 5 pm and patient discharged from recovery room before 7 pm
	Without SMOTE						With SMOTE
Classifier	Precision	Recall	MCC	Sensitivity	Specificity	AUC	Precision	Recall	MCC	Sensitivity	Specificity	AUC
Logistic regression	0.956	0.996	0.152	0.996	0.064	0.809	0.986	0.784	0.269	0.784	0.766	0.829
Balanced random forest classifier	0.991	0.855	0.378	0.855	0.832	0.913	0.979	0.974	0.516	0.984	0.564	0.919
Balanced bagging classifier	0.991	0.882	0.414	0.883	0.824	0.919	0.981	0.979	0.555	0.978	0.583	0.928
Random forest classifier	0.968	0.996	0.480	0.996	0.309	0.929	0.979	0.974	0.517	0.974	0.569	0.919
Multilayer perceptron neural network	0.959	0.994	0.231	0.996	0.123	0.844	0.984	0.838	0.295	0.838	0.711	0.847
Support vector classifier	0.956	0.998	0.149	0.998	0.049	0.724	0.844	0.849	0.327	0.848	0.751	0.858
2 pm start: predicting if surgery will end before 5 pm and patient discharged from recovery room before 7 pm
	Without SMOTE						With SMOTE
Classifier	Precision	Recall	MCC	Sensitivity	Specificity	AUC	Precision	Recall	MCC	Sensitivity	Specificity	AUC
Logistic regression	0.909	0.982	0.327	0.982	0.222	0.798	0.950	0.766	0.317	0.766	0.685	0.796
Balanced random forest classifier	0.972	0.832	0.476	0.832	0.811	0.899	0.952	0.952	0.576	0.952	0.625	0.903
Balanced bagging classifier	0.969	0.870	0.512	0.870	0.782	0.905	0.953	0.961	0.604	0.961	0.624	0.912
Random forest classifier	0.933	0.982	0.543	0.982	0.445	0.909	0.952	0.951	0.572	0.951	0.622	0.901
Multilayer perceptron neural network	0.912	0.980	0.352	0.980	0.253	0.824	0.953	0.800	0.358	0.800	0.691	0.828
Support vector classifier	0.914	0.981	0.381	0.981	0.275	0.769	0.961	0.814	0.405	0.814	0.739	0.840
3 pm start: predicting if surgery will end before 5 pm and patient discharged from recovery room before 7 pm
	Without SMOTE						With SMOTE
Classifier	Precision	Recall	MCC	Sensitivity	Specificity	AUC	Precision	Recall	MCC	Sensitivity	Specificity	AUC
Logistic regression	0.888	0.970	0.433	0.970	0.353	0.822	0.935	0.780	0.394	0.780	0.713	0.816
Balanced random forest classifier	0.963	0.815	0.524	0.815	0.835	0.910	0.937	0.941	0.609	0.941	0.661	0.910
Balanced bagging classifier	0.961	0.853	0.563	0.853	0.816	0.916	0.936	0.952	0.632	0.952	0.656	0.919
Random forest classifier	0.918	0.975	0.608	0.975	0.539	0.918	0.937	0.941	0.611	0.941	0.664	0.910
Multilayer perceptron neural network	0.891	0.971	0.448	0.971	0.365	0.847	0.938	0.805	0.427	0.805	0.716	0.846
Support vector classifier	0.886	0.980	0.447	0.980	0.329	0.809	0.952	0.783	0.453	0.783	0.789	0.855
4 pm start: predicting if surgery will end before 5 pm and patient discharged from recovery room before 7 pm
	Without SMOTE						With SMOTE
Classifier	Precision	Recall	MCC	Sensitivity	Specificity	AUC	Precision	Recall	MCC	Sensitivity	Specificity	AUC
Logistic regression	0.674	0.588	0.449	0.588	0.846	0.821	0.736	0.764	0.465	0.764	0.720	0.809
Balanced random forest classifier	0.701	0.833	0.620	0.833	0.808	0.900	0.829	0.775	0.632	0.775	0.861	0.901
Balanced bagging classifier	0.726	0.817	0.634	0.817	0.832	0.903	0.837	0.773	0.642	0.773	0.871	0.905
Random forest classifier	0.779	0.732	0.629	0.732	0.887	0.902	0.829	0.774	0.629	0.774	0.860	0.901
Multilayer perceptron neural network	0.699	0.640	0.501	0.640	0.849	0.843	0.771	0.794	0.523	0.794	0.749	0.849
Support vector classifier	0.708	0.670	0.527	0.670	0.850	0.844	0.767	0.817	0.534	0.817	0.740	0.846

Abbreviations: AUC, area under the receiver operating characteristic curve; MCC, Matthews correlation coefficient; PACU, postanesthesia care unit; SMOTE, synthetic minority oversampling technique.