2

Actual
diabetes
density

N
Y

0.00 0.25 0.50 0.75 1.00

Predicted Probability of diabetes
AUPR = 0.72, AUROC = 0.85
 Predicted N Predicted Y

Actual
diabetes
density

N
Y

0.00 0.25 0.50 0.75 1.00

Predicted Probability of diabetes
AUPR = 0.72, AUROC = 0.85
 insulin ●

skinfold ●

diastolic_bp ●

weight_class ●

plasma_glucose ●

pregnancies ●

pedigree ●

patient_id ●

diabetes ●

age ●

0% 10% 20% 30% 40% 50%

Percent of Observations Missing
 Random Forest
mtry
● ●

● ●

6 ● ●

● ● ● ● ● ●

3 ● ●

● ● ● ●

● ●

splitrule

extratrees ● ● ● ●● ● ● ● ●●

gini ● ● ● ● ● ● ● ● ● ● ● ●

min.node.size
20 ●
●
●

● ●
●
● ●
15
●
●
●
10 ●
●
● ●●
●
5 ●
●

● ●

0.65 0.66 0.67 0.68 0.69 0.70

AUPR
 Coefficients for regularized classification model of diabete

plasma_glucose ●

pregnancies ●

pedigree ●

weight_class_morbidly.obese (vs. obese) ●

age ●

skinfold ●

insulin ●

diastolic_bp ●

weight_class_missing (vs. obese) ●

weight_class_overweight (vs. obese) ●

weight_class_normal (vs. obese) ●

weight_class_underweight (vs. obese) ●

−1 0 1
Coefficient Estimate
Hyperparameter values alpha = 0 and lambda = 0.0168
 Random Forest variable importance

plasma_glucose ●

age ●

pregnancies ●

pedigree ●

insulin ●

skinfold ●

diastolic_bp ●

weight_class_normal ●

weight_class_overweight ●

weight_class_morbidly.obese ●

weight_class_missing ●

weight_class_underweight ●

0 25 50 75 100
Relative Importance
 0.8

0.6

weight_class
predicted_diabetes

morbidly obese
obese

0.4 NA
overweight
normal

0.2

0.0
75 100 125 150 175
plasma_glucose
Predictions made by: glmnet
 2.5 low extreme
moderate high

2.0

1.5
Actual
diabetes
density

N
Y

1.0

0.5

0.0

0.00 0.25 0.50 0.75 1.00

Predicted Probability of diabetes
AUPR = 0.7, AUROC = 0.88