Fig. 2
From: VENturing into machine learning for the morphological analysis of von Economo neurons
Most important variables found in human and information driven variable selection. The median ranking is represented by the x values and the numbers assigned to each point. A lower value indicates a higher importance of the variable. For example, the most important variables for all machine learning classifiers were average length and number of stems, but this level of agreement was not observed between the human experts, for whom the variable with the best average ranking was total number of trees, which appeared in seventh place for the machine learning classifiers. The interrater agreement between experts’ response was low (Fleiss’ Kappa for m raters; number of variables = 19, experts = 7, Kappa = 0.0132, z = 1.12, p = 0.262). Spearman rank correlation between human-driven and information-driven rankings: ρ = 0.28, p = 0.24 (indicating low agreement between methods).
