The previous section showed that benchmarking results in subgroups may be considerably different from that of the entire datasets collection. Going one step further, one can extend the analysis of meta-features towards meta-learning to gain insight on their influence. More precisely, taking the datasets as observations we build a regression RF that predicts the difference in performance between RF and LR based on the four meta-features considered in the previous subsection \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\left ({p}, {n}, \frac {p}{n} \text { and } C_{max}\right)$\end{document}p,n,pnandCmax. Figure 6 depicts partial dependence plots for visualization of the influence of each meta-feature. Again, we notice a dependency on p and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\frac {p}{n}$\end{document}pn as outlined in “Subgroup analyses: meta-features” section and the comparatively bad results of RF when compared to LR for datasets with small p. The importance of Cmax and n is less noticeable.
