Does multidimensional daily information predict the onset of myopia? A 1-year prospective cohort study

Table 3 Model performance using five algorithms in test set

Model	Precision	Recall	F1-score	AUC
LR	0.892	0.014	0.027	0.739
SVM	0.883	0.013	0.025	0.647
GBDT	0.887	0.221	0.336	0.865
RF	0.934	0.494	0.651	0.935
CB	0.953	0.639	0.774	0.951
CB (without the SMOTE)	0.889	0.133	0.213	0.763
CB (the compact model)	0.905	0.320	0.432	0.891
LR (the compact model)	0.887	0.133	0.211	0.692

The 10 predictors used on the compact model: parental myopia, grade, frequency of feeling eye fatigue, height, weight, frequency of visual health education from parents, academic level, number of after-school tutoring per week, frequency of fish intake in the diet and hours of outdoor activities per day on school days
LR Logistic Regression, SVM Support Vector Machines, RF Random Forest, GBDT Gradient Boosting Decision Tree, CB CatBoost

ISSN: 1475-925X