Simple Method to Predict Insulin Resistance in Children Aged 6–12 Years by Using Machine Learning

Figures & data

Table 1 Grid Search Parameter Intervals for Model Training

Model	Parameter 1	Parameter 2	Parameter 3	Parameter 4	Parameter 5
RF	n_estimators: range (30, 80, 10)	Max_depth: range (3, 10, 2)	Min_samples_leaf: [5, 6, 7]	Max_features: [1, 2, 3]
LR	Max_iter: [20, 40, 60, 100]	C: [0.01, 0.1, 1, 10]
SVM	Kernel: [linear, poly, rbf]	C: [1, 10, 100]	Gamma: [1, 0.1, 0.01, 0.001]
XGBoost	n_estimators: range (80, 200, 20)	Max_depth: range (2, 15, 2)	Subsample: np.linspace (0.7, 0.9, 3)	Colsample_bytree: np.linspace (0.5, 0.98, 4)	Min_child_weight: range (1, 9, 3)
CatBoost	Depth: range (4, 10, 1)	Learning_rate: [0.05, 0.1, 0.15]	l2_leaf_reg: [1, 4, 9]

Abbreviations: LR, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting; CatBoost, gradient boosting with categorical features support.

Figure 1 Flowchart of the model training and test procedure in this study.

Abbreviations: SMOTE, Synthetic Minority Oversampling Technique; RFE, recursive feature elimination; LR, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting; CatBoost, gradient boosting with categorical features support; AUC, area under the curve.

Figure 2 Distribution of missing features in the training set and test set.

Abbreviations: HGB, haemoglobin; WBC, leukocytes; RBC, erythrocytes; PLT, platelets; SBP, systolic blood pressure; DBP, diastolic blood pressure; BMI, body mass index.

Table 2 Feature Comparison Between the Training Set and Test Set- Continuous Variables (Mean± SD)

	Training Set	Test Set	t	p
Age	9.64±1.74	9.40±1.75	1.351	0.177
BMI (kg/m^2)	16.69±3.04	24.63±3.40	−24.112	<0.001
SBP (mmHg)	97.43±12.62	113.93±11.50	−13.462	<0.001
DBP (mmHg)	64.98±9.00	64.36±8.28	0.715	0.475
Hip circumference (cm)	71.40±9.87	99.24±10.77	−27.967	<0.001
Waist circumference (cm)	59.51±8.93	82.57±10.82	−22.289	<0.001
Sleep duration (hours/day)	9.91±1.01	9.13±0.80	0.688	0.492
Sports activities in school per week (times/week)	7.71±5.55	4.24±1.97	11.395	<0.001
HGB (g/L)	134.30±15.15	137.83±8.78	−3.423	0.001
WBC (*10^9/L)	6.75±1.85	8.38±2.07	−8.106	<0.001
RBC (*10^9/L)	4.79±0.56	5.02±0.33	−6.007	<0.001
PLT (*10^9/L)	280.44±78.09	309.60±62.05	−4.487	<0.001
Glucose (mmol/L)	4.89±0.82	5.02±0.41	−1.805	0.072

Note: p values <0.05 were considered significant.

Abbreviations: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; HGB, haemoglobin; WBC, leukocytes; RBC, erythrocytes; PLT, platelets.

Table 3 Feature Comparison Between the Training Set and Test Set - Categorical Variables

	Training Set	Test Set	χ^2	p
Males, n, (%)	271(55.1%)	78(60.5%)	1.203	0.273
Children with IR, n, (%)	155(31.6%)	96(74.4%)	95.891	<0.001
Outdoor sports activities, n, (%)	174(55.1%)	89(60.5%)	47.333	<0.001
Sports activities in school, n, (%)	425(86.4%)	103(79.8%)	3.043	0.064

Note: p values <0.05 were considered significant.

Figure 3 Violin plot depicting the comparison of basic information between the training set and the external test set.

Abbreviations: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure.

Table 4 Features Selected by Different Models, Displayed in Descending Order of Importance Percentage

Rank	LR	SVM	RF	XGBoost	CatBoost
1	Glucose (44.69%)	Glucose (42.23%)	Glucose (16.62%)	Age (17.31%)	PLT (18.87%)
2	Age (17.63%)	Gender (16.27%)	Hip circumference (12.77%)	Hip circumference (8.82%)	SBP (18.47%)
3	Sports activities in school (16.80%)	Age (15.73%)	DBP (12.49%)	Glucose (8.72%)	Glucoses (17.49%)
4	Gender (15.29%)	Sports activities in school (12.96%)	Waist circumference (12.30%)	Waist circumference (8.04%)	HGB (16.62%)
5	WBC (3.89%)	Outdoor sports activities (6.74%)	SBP (9.26%)	BMI (7.94%)	Hip circumference (15.40%)
6	Waist circumference (1.70%)	WBC (1.85%)	WBC (8.31%)	HGB (6.90%)	Waist circumference (13.16%)
7		DBP (1.69%)	HGB (7.86%)	WBC (6.16%)
8		Waist circumference (1.33%)	RBC (6.99%)	PLT (5.95%)
9		RBC (1.21%)	PLT (6.86%)	Gender (5.73%)
10			Sports activities in school per week (6.54%)	Sports activities in school (5.46%)
11				RBC (5.31%)
12				DBP (5.06%)
13				Sports activities in school per week (4.57%)
14				SBP (4.03%)

Abbreviations: LR, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting; CatBoost, gradient boosting with categorical features support; HGB, haemoglobin; WBC, leukocytes; RBC, erythrocytes; PLT, platelets; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure.

Figure 4 Feature selection for different models by the RFE method. (A) LR; (B) RF; (C) SVM; (D) XGBoost; (E) CatBoost.

Abbreviations: RFE, recursive feature elimination; LR, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting; CatBoost, gradient boosting with categorical features support.

Figure 5 ROC curves for the test set.

Abbreviations: ROC, receiver operating characteristic; AUC, area under the curve; LR, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting; CatBoost, gradient boosting with categorical features support.

Table 5 Evaluation Metrics for the Different Models

	Accuracy	Sensitivity	Specificity	Precision	F1 Score
RF	0.81	0.88	0.61	0.87	0.87
LR	0.80	0.89	0.55	0.85	0.87
SVM	0.74	0.77	0.67	0.87	0.82
XGBoost	0.78	0.78	0.79	0.91	0.84
CatBoost	0.77	0.79	0.69	0.88	0.84

Abbreviations: LR, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting; CatBoost, gradient boosting with categorical features support.