Figure 11.
The image comprises five scatter plots labelled, a, to, e, each illustrating the relationship between predicted values and test values, both measured on the horizontal and vertical axes respectively, with values denoted in units of 10 to the power of negative 6. Each plot presents a set of blue data points, amounting to 1650 for each graph, and includes black regression lines. Additionally, two dashed red lines in each plot indicate the prediction intervals, labelled with percentages that represent the deviation from the regression line. The coefficient of determination, R squared, and normalised root mean square error, N R M S E, are also detailed within each subplot, illustrating the varying accuracies of the predictions across the different plots. Each subplot follows a consistent layout for ease of comparison.Evaluation of ML methods on validation data of total shrinkage: (a) DT; (b) RF; (c) ET; (d) GB; (e) XGBoost