fig4

Figure 4. Data sampling and model construction for the Ma-TE dataset. (A) Exploration of MLMD-TE data subsets and algorithm combinations, including the original baseline model’s 10-fold R² value; (B) Presentation of error performance for six algorithms on the validation and test sets; (C) Comparison of true and predicted values for the best generalization model XGBR on the training and validation sets; (D) Comparison of true and predicted values for the XGBR model on the test set. R²: Determination coefficient; XGBR: XGBoost regression.