Machine Learning-Based Prediction Model of Preterm Birth Using Electronic Health Record

Table 3

The performance of models in the test set.

	Models	Accuracy	AUC (95% CI)	Sensitivity	Specificity
Dataset 1	SVM	0.720	0.791 (0.771–0.811)	0.710	0.731
	RF	0.777	0.861 (0.841–0.871)	0.720	0.840
	NBM	0.677	0.741 (0.721–0.761)	0.705	0.646
	ANN	0.634	0.691 (0.671–0.711)	0.687	0.576
	K-means	0.611	0.681 (0.661–0.701)	0.794	0.412
	Log	0.610	0.701 (0.681–0.721)	0.378	0.861
Dataset 2	SVM	0.721	0.791 (0.781–0.811)	0.722	0.721
	RF	0.794	0.871 (0.851–0.881)	0.756	0.832
	NBM	0.682	0.771 (0.751–0.791)	0.785	0.581
	ANN	0.666	0.731 (0.711–0.751)	0.595	0.738
	K-means	0.602	0.681 (0.671–0.701)	0.811	0.393
	Log	0.606	0.701 (0.681–0.721)	0.364	0.847
Dataset 3	SVM	0.719	0.801 (0.781–0.811)	0.695	0.743
	RF	0.806	0.881 (0.871–0.901)	0.765	0.846
	NBM	0.674	0.791 (0.771–0.811)	0.837	0.515
	ANN	0.733	0.801 (0.791–0.821)	0.741	0.726
	K-means	0.612	0.711 (0.691–0.731)	0.824	0.405
	Log	0.633	0.701 (0.681–0.721)	0.421	0.839
Dataset 4	SVM	0.719	0.791 (0.781–0.811)	0.678	0.763
	RF	0.807	0.881 (0.871–0.891)	0.743	0.875
	NBM	0.626	0.741 (0.721–0.761)	0.328	0.946
	ANN	0.732	0.811 (0.801–0.831)	0.730	0.734
	K-means	0.626	0.721 (0.701–0.741)	0.801	0.436
	Log	0.611	0.701 (0.691–0.721)	0.361	0.880
Dataset 5	SVM	0.729	0.801 (0.781–0.811)	0.685	0.773
	RF	0.816	0.891 (0.871–0.901)	0.751	0.882
	NBM	0.622	0.741 (0.721–0.761)	0.315	0.937
	ANN	0.747	0.811 (0.801–0.831)	0.730	0.763
	K-means	0.609	0.701 (0.681–0.721)	0.780	0.434
	Log	0.623	0.691 (0.671–0.711)	0.391	0.861

NBM: Naive Bayesian; SVM: Support Vector Machine; RF: Random Forest Tree; ANN: Artificial Neural Networks; Log: Logistic regression; Dataset 1: 20 weeks gestation; Dataset 2: 22 weeks gestation; Dataset 3: 24 weeks gestation; Dataset 4: 26 weeks gestation; Dataset 5: 27 weeks gestation. AUC: the area under the curve; CI: confidence interval.