Fig. 3
Construction of NASH fibrosis diagnosis model based on machine learning model. (A) The heatmap illustrates the area under the curve (AUC) of the performance metrics for both the training and validation sets, evaluated across 12 machine learning models and 103 algorithms. The variable selection and model-building algorithms were used in the order shown in the text. (B) The ROC curves demonstrate the testing performance of the model based on the RF (Random Forest) + Enet (Elastic Net, α = 0.6) algorithm for both the training and validation sets. (C) The confusion matrix reveals the sensitivity and specificity of the diagnostic model in detecting NASH fibrosis within the training and validation sets.
