Table 4.

Top five factors for best performing models to predict food intakes at eating occasions and daily diet quality, using mean absolute SHAP values

Outcome	Occurrence		Amount
	Feature	Mean SHAP	Feature	Mean SHAP
Fruits*	-	-	Self-efficacy	0.0005
	-	-	Food availability	0.0005
	-	-	Cooking confidence	0.0005
	-	-	Social support from family	0.0005
	-	-	Food choice barriers	0.0005
Vegetables*	-	-	SEIFA	0.0035
	-	-	Income	0.0035
	-	-	Social support from family	0.0034
	-	-	Proximity and access to food destinations	0.0034
	-	-	Age	0.0033
Grains	Activity at consumption	0.0030	Food choice barriers	0.0008
	Education	0.0024	Living situation	0.0008
	Country of birth	0.0023	Age	0.0007
	Food availability	0.0023	Social support from friends/colleagues	0.0007
	Place of purchase	0.0022	Perceived time scarcity	0.0007
Meat and alternatives	Cooking confidence	0.0010	Smoking status	0.0014
	Food choice barriers	0.0009	Place of purchase	0.0014
	Proximity and access to food destinations	0.0009	Food availability	0.0013
	education	0.0008	Food choice barriers	0.0013
	Country of birth	0.0008	Income	0.0013
Dairy and alternatives*	-	-	Education	0.0012
	-	-	PA	0.0011
	-	-	Country of birth	0.0011
	-	-	Food choice barriers	0.0011
	-	-	Food availability	0.0011
Discretionary foods	Place of purchase	0.0024	Place of consumption	0.0037
	Social support from friends/colleagues	0.0023	Perceived time scarcity	0.0037
	Education	0.0022	Social support from family	0.0036
	Proximity and access to food destinations	0.0021	Age	0.0036
	Country of birth	0.0019	Proximity and access to food destinations	0.0036
Daily diet quality*	-	-	Cooking confidence	0.0814
	-	-	Self-efficacy	0.0800
	-	-	Food availability	0.0772
	-	-	Perceived time scarcity	0.0760
	-	-	Activity at consumption	0.0753

*The best performing for these outcomes are random forest model. They do not have two parts (occurrence and amount) like Hurdle models