fig5
Figure 5. Comparison of testing R2, NRMSE and MRE among different ML algorithms from varying feature sets: (A-C) NoGPF; (D-F) GPFOnly; and (G-I) GPFCombined. The results came from ten times of holdouts. R2: The coefficient of determination; NRMSE: normalized root mean squared error; MRE: mean relative error; ML: machine learning.
