Fig. 6: Comparison between experimental and ML predicted corrosion rate.

a Using top 13 features from stage 1 down selection, and b using 5 optimized features from stage 2 down selection. The bad prediction for each case is indicated by blue arrows.
a Using top 13 features from stage 1 down selection, and b using 5 optimized features from stage 2 down selection. The bad prediction for each case is indicated by blue arrows.