Automated chest screening based on a hybrid model of transfer learning and convolutional sparse denoising autoencoder

Table 5 Performance results based on test data using four classifiers: KNN, logistic regression, SVM, random forest

Classifier	Data augmentation	Precision	Recall	F1	AUC
KNN	Without augmentation	0.73	0.58	0.64	0.753
	Without augmentation	0.87	0.93	0.9	0.753
	SMOTE	0.44	0.84	0.58	0.752
	SMOTE	0.93	0.67	0.78	0.752
	Positive augmentation	0.57	0.68	0.62	0.758
	Positive augmentation	0.89	0.68	0.87	0.758
	4× augmentation	0.57	0.84	0.62	0.758
	4× augmentation	0.89	0.89	0.87	0.758
Logistic	Without augmentation	0.76	0.54	0.63	0.744
	Without augmentation	0.87	0.95	0.91	0.744
	SMOTE	0.61	0.82	0.64	0.79
	SMOTE	0.92	0.76	0.84	0.79
	Positive augmentation	0.57	0.66	0.61	0.754
	Positive augmentation	0.89	0.84	0.86	0.754
	4× augmentation	0.63	0.72	0.67	0.792
	4× augmentation	0.91	0.87	0.89	0.792
SVM	Without augmentation	0.63	0.73	0.68	0.798
	Without augmentation	0.91	0.86	0.89	0.798
	SMOTE	0.63	0.77	0.69	0.813
	SMOTE	0.92	0.86	0.89	0.813
	Positive augmentation	0.59	0.64	0.61	0.75
	Positive augmentation	0.88	0.86	0.87	0.75
	4× augmentation	0.61	0.71	0.66	0.784
	4× augmentation	0.9	0.86	0.88	0.784
Random forest	Without augmentation	0.69	0.38	0.49	0.663
	Without augmentation	0.83	0.95	0.88	0.663
	SMOTE	0.62	0.5	0.56	0.704
	SMOTE	0.85	0.9	0.88	0.704
	Positive augmentation	0.54	0.63	0.58	0.73
	Positive augmentation	0.88	0.83	0.85	0.73
	4× augmentation	0.59	0.54	0.56	0.71
	4× augmentation	0.86	0.88	0.87	0.71

ISSN: 1475-925X