Table 3.

The performance of each model including the two null models.

Prediction model	µ-averaged accuracy (%)	µ-averaged F1-score	AUC	RMSE
Null model 1: random	19.7	0.20	0.5	2.2
Null model 2: mode	26.7	0.27	0.5	1.6
Multinomial regression model	84.6	0.31	0.7	1.6
Gradient boosting model	87.0	0.41	0.7	1.5
Random forest model	91.9	0.63	0.9	1.1

µ, micro; AUC, area under the receiving operating characteristic curve; RMSE, root-mean-square error.