Table 1. Parameter values used in iterative optimization of classification models and final values used for each parameter for each country-specific model.

Parameter	Definition	Range Tested	Increments	Kenya	Tanzania
eta	Learning rate: shrinks feature weights after each round to reach the best optimum	0.02–0.3	0.02	0.18	0.28
max_depth	Maximum tree depth: model complexity	2–6	1	6	5
subsample	Subsamples: the number of observations supplied to a tree	0.25–0.75	0.25	0.50	0.25
colsample_bytree	Feature samples: the number of features supplied to a tree	1/3–2/3	1/3	2/3	2/3
scale_pos_weight	Positive weight scale: corrects for inbalances in response variable values	1.2–5	0.2	1.6	1.2