Joint modeling strategy for using electronic medical records data to build machine learning models: an example of intracerebral hemorrhage

Tang, Jianxiang; Wang, Xiaoyu; Wan, Hongli; Lin, Chunying; Shao, Zilun; Chang, Yang; Wang, Hexuan; Wu, Yi; Zhang, Tao; Du, Yu

doi:10.1186/s12911-022-02018-x

BMC Medical Informatics and Decision Making

Table 7 The performance of all 84 combinations of joint modeling strategy

From: Joint modeling strategy for using electronic medical records data to build machine learning models: an example of intracerebral hemorrhage

CVCF	Resampling	Models	Accuracy	Sensitivity	Specificity	AUC	Precision	F1
No	Original	LR	0.787	0.424	0.948	0.733	0.784	0.550
No	Original	RF	0.811	0.426	0.983	0.743	0.920	0.581
No	Original	ANN	0.809	0.415	0.985	0.751	0.925	0.572
No	Original	SVM	0.807	0.420	0.979	0.725	0.903	0.571
No	Original	KNN	0.798	0.395	0.977	0.737	0.886	0.545
No	Original	Stacking	0.813	0.435	0.981	0.756	0.916	0.588
No	Original	AdaBoost	0.816	0.455	0.977	0.743	0.897	0.602
No	RUS	LR	0.777	0.518	0.893	0.738	0.684	0.589
No	RUS	RF	0.789	0.498	0.919	0.750	0.736	0.592
No	RUS	ANN	0.784	0.504	0.909	0.739	0.714	0.590
No	RUS	SVM	0.766	0.547	0.864	0.728	0.644	0.591
No	RUS	KNN	0.785	0.485	0.920	0.744	0.730	0.581
No	RUS	Stacking	0.773	0.533	0.881	0.745	0.669	0.591
No	RUS	AdaBoost	0.786	0.496	0.915	0.738	0.727	0.587
No	ROS	LR	0.774	0.527	0.884	0.736	0.671	0.589
No	ROS	RF	0.784	0.513	0.906	0.750	0.708	0.594
No	ROS	ANN	0.778	0.537	0.886	0.750	0.681	0.599
No	ROS	SVM	0.751	0.572	0.830	0.740	0.603	0.586
No	ROS	KNN	0.775	0.458	0.917	0.733	0.713	0.555
No	ROS	Stacking	0.767	0.557	0.860	0.752	0.642	0.596
No	ROS	AdaBoost	0.778	0.514	0.897	0.740	0.690	0.588
No	ADASYN	LR	0.761	0.516	0.870	0.722	0.640	0.570
No	ADASYN	RF	0.757	0.523	0.862	0.737	0.629	0.569
No	ADASYN	ANN	0.740	0.531	0.834	0.711	0.591	0.557
No	ADASYN	SVM	0.726	0.564	0.798	0.718	0.556	0.558
No	ADASYN	KNN	0.748	0.514	0.853	0.706	0.610	0.556
No	ADASYN	Stacking	0.749	0.548	0.839	0.718	0.605	0.574
No	ADASYN	AdaBoost	0.751	0.523	0.853	0.706	0.615	0.564
No	BSMOTE*	LR	0.729	0.584	0.794	0.732	0.562	0.571
No	BSMOTE	RF	0.720	0.570	0.788	0.737	0.549	0.558
No	BSMOTE	ANN	0.720	0.564	0.791	0.724	0.548	0.555
No	BSMOTE	SVM	0.701	0.584	0.755	0.720	0.520	0.547
No	BSMOTE	KNN	0.707	0.563	0.771	0.718	0.529	0.543
No	BSMOTE	Stacking	0.710	0.602	0.758	0.733	0.529	0.561
No	BSMOTE	AdaBoost	0.713	0.572	0.776	0.721	0.538	0.552
No	SMOTEENN	LR	0.680	0.603	0.715	0.718	0.497	0.539
No	SMOTEENN	RF	0.651	0.646	0.654	0.696	0.459	0.534
No	SMOTEENN	ANN	0.652	0.642	0.656	0.706	0.459	0.533
No	SMOTEENN	SVM	0.661	0.645	0.669	0.707	0.474	0.542
No	SMOTEENN	KNN	0.683	0.570	0.735	0.686	0.501	0.526
No	SMOTEENN	Stacking	0.638	0.662	0.628	0.672	0.449	0.531
No	SMOTEENN	AdaBoost	0.645	0.654	0.642	0.703	0.460	0.535
Yes	Original	LR	0.797	0.393	0.977	0.722	0.885	0.543
Yes	Original	RF	0.812	0.424	0.986	0.737	0.933	0.581
Yes	Original	ANN	0.816	0.436	0.985	0.745	0.931	0.592
Yes	Original	SVM	0.816	0.436	0.986	0.746	0.938	0.593
Yes	Original	KNN	0.796	0.368	0.987	0.724	0.924	0.525
Yes	Original	Stacking	0.816	0.436	0.986	0.725	0.933	0.593
Yes	Original	AdaBoost	0.814	0.431	0.985	0.747	0.927	0.587
Yes	RUS	LR	0.794	0.468	0.939	0.738	0.775	0.582
Yes	RUS	RF	0.814	0.433	0.984	0.748	0.925	0.588
Yes	RUS	ANN	0.815	0.437	0.984	0.742	0.928	0.593
Yes	RUS	SVM	0.815	0.437	0.984	0.746	0.928	0.592
Yes	RUS	KNN	0.805	0.453	0.963	0.725	0.845	0.589
Yes	RUS	Stacking	0.816	0.435	0.986	0.724	0.933	0.592
Yes	RUS	AdaBoost	0.814	0.433	0.984	0.734	0.923	0.588
Yes	ROS	LR	0.795	0.471	0.939	0.735	0.777	0.585
Yes	ROS	RF	0.809	0.440	0.974	0.752	0.889	0.587
Yes	ROS	ANN	0.814	0.440	0.981	0.749	0.918	0.593
Yes	ROS	SVM	0.811	0.451	0.972	0.748	0.883	0.594
Yes	ROS	KNN	0.786	0.445	0.938	0.734	0.764	0.561
Yes	ROS	Stacking	0.813	0.440	0.980	0.742	0.911	0.591
Yes	ROS	AdaBoost	0.810	0.443	0.974	0.746	0.888	0.589
Yes	ADASYN	LR	0.776	0.424	0.934	0.736	0.741	0.538
Yes	ADASYN	RF	0.813	0.437	0.981	0.738	0.917	0.590
Yes	ADASYN	ANN	0.814	0.437	0.982	0.752	0.919	0.590
Yes	ADASYN	SVM	0.813	0.437	0.981	0.746	0.917	0.590
Yes	ADASYN	KNN	0.806	0.446	0.966	0.724	0.856	0.585
Yes	ADASYN	Stacking	0.814	0.436	0.983	0.714	0.923	0.590
Yes	ADASYN	AdaBoost	0.813	0.436	0.981	0.732	0.915	0.589
Yes	BSMOTE	LR	0.773	0.445	0.919	0.736	0.714	0.546
Yes	BSMOTE	RF	0.813	0.433	0.983	0.743	0.922	0.588
Yes	BSMOTE	ANN	0.815	0.441	0.982	0.742	0.918	0.594
Yes	BSMOTE	SVM	0.816	0.439	0.984	0.749	0.927	0.594
Yes	BSMOTE	KNN	0.804	0.429	0.971	0.723	0.872	0.574
Yes	BSMOTE	Stacking	0.815	0.438	0.984	0.727	0.926	0.593
Yes	BSMOTE	AdaBoost	0.813	0.436	0.982	0.738	0.919	0.589
Yes	SMOTEENN	LR	0.795	0.466	0.942	0.735	0.783	0.582
Yes	SMOTEENN	RF	0.814	0.435	0.983	0.746	0.922	0.589
Yes	SMOTEENN	ANN	0.814	0.440	0.981	0.743	0.914	0.593
Yes	SMOTEENN	SVM	0.816	0.445	0.982	0.747	0.921	0.598
Yes	SMOTEENN	KNN	0.805	0.438	0.968	0.724	0.862	0.580
Yes	SMOTEENN	Stacking	0.816	0.444	0.982	0.711	0.920	0.597
Yes	SMOTEENN	AdaBoost	0.815	0.437	0.983	0.735	0.924	0.591

* BSMOTE: Borderline SMOTE; Bold indicates the best value

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com