Page 38 - Read Online
P. 38
Lu et al. J Mater Inf 2024;4:31 https://dx.doi.org/10.20517/jmi.2024.65 Page 11 of 16
Figure 7. Workflow of the machine learning approach.
2
Figure 8. (A-C) Comparison of DFT-calculated ΔG values with those predicted by three ML algorithms; (D) Average R and MAE
scores from 5-fold cross-validation for three ML models. DFT: Density functional theory; ML: Machine learning; MAE: Mean absolute
error.
protonation steps, and identifying the side-on and end-on adsorption modes in the first protonation step.
The Mendeleev number (N ) and N of the metal atom ranked as the second and third most important
d
m
features, contributing 23.77% and 8.94%, respectively (see Figure 9A). This indicates that these two intrinsic
