Deep Learning-Based Prediction of Physical Stability considering Class Imbalance for Amorphous Solid Dispersions

Table 1

The performance comparison of models applying various sampling techniques to the proposed neural net architecture.

		Without sampling	Oversampling		Undersampling				Hybrid sampling
			SMOTE	ADASYN	TL	ENN	OSS	CNN	Smote+ENN	Smote+TL
Accuracy	Training set	0.864	0.998	0.99	1.0	1.0	1.0	1.0	1.0	1.0
	5-fold cross-validation
	Test set	0.85	0.9	0.8	0.9	0.8	0.8	0.65	0.95	0.9
Precision	Training set	0.865	0.998	0.99	1.0	1.0	1.0	1.0	1.0	1.0
	5-fold cross-validation
	Test set	0.864	0.86	0.864	0.86	0.887	0.826	0.725	0.96	0.86
Recall	Training set	0.864	0.998	0.99	1.0	1.0	1.0	1.0	1.0	1.0
	5-fold cross-validation
	Test set	0.85	0.9	0.8	0.9	0.8	0.8	0.65	0.95	0.9
-score	Training set	0.86	0.998	0.99	1.0	1.0	1.0	1.0	1.0	0.998
	5-fold cross-validation
	Test set	0.842	0.878	0.812	0.878	0.825	0.803	0.676	0.952	0.878