Skip to main content
. 2025 Aug 12;13:1537098. doi: 10.3389/fped.2025.1537098

Figure 1.

Flowchart depicting a model training pipeline for imbalance mitigation in fibrosis classification. It includes steps for class distribution analysis, imbalance mitigation strategies like class weight adjustment and SMOTE implementation, and stratified data splitting. The pipeline tests algorithms such as weighted logistic regression, random forest, and CatBoost, with validation protocols like five-fold cross-validation. Metrics tracked include recall, F1-score, and ROC-AUC. Optimal strategy selection leads to outcomes like the final model and clinical interpretation. A legend clarifies sections like data assessment, techniques, parameters, and models.

This figure depicts a machine learning pipeline for predicting pediatric NAFLD, starting with data assessment and class adjustment using weighted techniques and SMOTE. It progresses through stratified data splitting, model training with algorithms like CatBoost, and validation using 5-fold cross-validation, tracking metrics like F1-score and ROC-AUC. The process culminates in optimal strategy selection and clinical interpretation for screening and biopsy guidance. Diagram created with MermaidChart (https://www.mermaidchart.com/).