Figure 5
From: The cutaneous microbiome in hospitalized patients with pressure ulcers
Random Forests with Boruta feature selection using species abundance data. (A) The importance of each feature across 1000 repeats is shown compared with the worst, best, and mean randomly generated features. (B) Variable importance in the random forest. (C) Confusion matrix. (D) Effect of species abundance on expected pressure ulcer occurrence in the random forest model. Values on the vertical axis represent the expected probability of a pressure ulcer for a given species, after adjusting for all other predictors. Dashed red lines represent ±2 standard error. The red dots represent samples with corresponding inner lines on the x-axis representing the values of the abundancies. Expected probability based on random classification is 0.44. (E) Interaction plot.
