Development and validation of a meta-learner for combining statistical and machine learning prediction models in individuals with depression

Table 3 Permutation feature importance values of the meta-learners

Outcome	Meta-learner	Base-learner	Change of measures (%)
PHQ-9,^** measured by MAE^††	Linear regression	Ridge regression	24.78
	Linear regression	MLP^‡‡	0.30
	MLP	Ridge regression	31.70
	MLP	MLP	2.34
Dropout, measured by AUC^§§	Logistic regression	Ridge regression	0.83
	Logistic regression	MLP	5.21
	MLP	Ridge regression	0.42
	MLP	MLP	2.67

ISSN: 1471-244X