Table 2.

Performance metrics for the three considered models in the training and validation sets. LR, linear regression. RF, random forest. FFNN, feed forward neural network. MAE, mean average error. Correlation, Pearson product-moment correlation coefficient. P10/P30, fraction of predictions with an error within the 10%/30% threshold. Best results for each set are highlighted in bold.

	Training Set			Validation Set			Test Set
	LR	RF	FFNN	LR	RF	FFNN	LR	RF	FFNN
MAE	0.2210	0.1801	0.1956	0.1975	0.1894	0.1780	0.2515	0.2257	0.2035
Correlation	0.8402	0.8869	0.8632	0.8732	0.8871	0.9028	0.7922	0.7450	0.8563
P10	0.2635	0.3322	0.3099	0.2699	0.2828	0.2879	0.2530	0.2803	0.3136
P30	0.6901	0.7751	0.7391	0.7044	0.7189	0.7532	0.6394	0.6727	0.6955