Feature importance plots — plot_feature

Various plotting functions to visualize feature importances. `plot_feature_importances` creates bar plots of the feature importances across different models. `plot_feature_importance_pair` create pair plots, comparing the feature importances between all pairs of models. `plot_feature_importance_stability` creates boxplots of the distribution of feature importances across the many data perturbations for each model.

plot_feature_importance(
  data,
  feature_col = "Variable",
  importance_col = "Importance",
  model_col = "Method",
  max_features = 50,
  use_rankings = FALSE,
  use_facets = TRUE,
  interactive = FALSE,
  ...
)

plot_feature_importance_pair(
  data,
  feature_col = "Variable",
  importance_col = "Importance",
  model_col = "Method",
  use_rankings = FALSE,
  interactive = FALSE,
  ...
)

plot_feature_importance_stability(
  data,
  feature_col = "Variable",
  importance_col = "Importance",
  model_col = "Method",
  max_features = 50,
  use_rankings = FALSE,
  use_facets = TRUE,
  interactive = FALSE,
  ...
)

Arguments

data: A data frame with the feature importance information. At a minimum, this data frame should contain columns with the names specified by `feature_col`, `importance_col`, and `model_col`.
feature_col: Character string. Name of the column in `data` with the feature names.
importance_col: Character string. Name of the column in `data` with the feature importance scores.
model_col: Character string. Name of the column in `data` with the model names.
max_features: Maximum number of features to display in the plot.
use_rankings: Logical indicating whether or not to use the feature importance rankings, rather than the raw feature importance scores, for plotting.
use_facets: Logical indicating whether or not to use faceting to separate the feature importances by model.
interactive: Logical indicating whether or not to return an interactive plot.
...: Additional arguments to pass to `plot_pairs()`. Only used in `plot_feature_importance_pair()`.

Value

If `interactive = FALSE`, returns a ggplot object. If `interactive = TRUE`, returns an interactive plotly object.