Table 2 Statistics on the representative validation set.

No. of features	3			5			63			1260
Model Type	Tiny model			Small model			Medium model			Large model
Selection method	a	b	c	a	b	c	a	b	c
RMSE/pp	50	24	66	51	26	60	49	23	40	38
R²	0.56	0.94	0.53	0.54	0.94	0.95	0.58	0.94	0.94	0.95
Pearson’s r	0.75	0.97	0.73	0.73	0.97	0.97	0.76	0.97	0.76	0.97
p-value	0.14	0.01	0.16	0.16	0.01	0.01	0.14	0.01	0.08	0.01

Root mean squared errors (RMSE), coefficients of determination (R²), correlation coefficients (Pearson’s r) and p-values of the representative validation set predictions for all trained models by model type and feature selection method (a: ANOVA, b: RFE, c: random selection). 3,5- Dinitrobenzoic acid (compound No. 13) was omitted for calculation of the statistical values as its molecular features are in parts substantially outside of the domain covered by the training data.

Quick links

Search