Table 3 Model comparisons based on the five machine leaning techniques.

Feature set	Machine learning algorithm	Validation method	N folds	Training set size	Testing set size	Training accuracy	Testing accuracy
1	DT	Holdout sampling		1,140	286	0.668	0.703
	DT	Cross-validation	5	912	286	0.625	0.692
	LR	Holdout sampling		1,140	286	0.663	0.647
	LR	Cross-validation	5	912	286	0.657	0.632
	Bagging	Holdout sampling		1,140	286	0.680	0.710
	Bagging	Cross-validation	5	912	286	0.655	0.706
	RF	Holdout sampling		1,140	286	0.675	0.713
	RF	Cross-validation	5	912	286	0.675	0.692
	AdaBoost	Holdout sampling		1,140	286	0.668	0.696
	Real AdaBoost	Cross-validation	5	912	286	0.642	0.713
2	DT	Holdout sampling		1,140	286	0.780	0.762
	DT	Cross-validation	5	912	286	0.758	0.745
	LR	Holdout sampling		1,140	286	0.791	0.746
	LR	Cross-validation	5	912	286	0.814	0.825
	Bagging	Holdout sampling		1,140	286	0.976	0.930
	Bagging	Cross-validation	5	912	286	0.794	0.776
	RF	Holdout sampling		1,140	286	0.949	0.916
	RF	Cross-validation	5	912	286	0.918	0.941
	AdaBoost	Holdout sampling		1,140	286	0.943	0.878
	Real AdaBoost	Cross-validation	5	912	286	0.932	0.948

DT, decision tree; RF, random forest; LR, logistic regression.
Feature set 1 includes sociodemographic and clinical variables.
Feature set 2 includes PRO variables and the variables included in feature set 1.

Quick links

Search