Plot ROC/PR curves for feature selection.

Plot ROC/PR curves for feature selection or some summary thereof across experimental replicates.

Usage

plot_feature_selection_curve(
  fit_results = NULL,
  eval_results = NULL,
  eval_name = NULL,
  eval_fun = "summarize_feature_selection_curve",
  eval_fun_options = NULL,
  vary_params = NULL,
  curve = c("ROC", "PR"),
  show = c("line", "ribbon"),
  ...
)

Arguments

fit_results

A tibble, as returned by the fit method.

eval_results

A list of result tibbles, as returned by the evaluate method.

eval_name

Name of Evaluator containing results to plot. If NULL, the data used for plotting is computed from scratch via eval_fun.

eval_fun

Character string, specifying the function used to compute the data used for plotting if eval_name = NULL. If eval_name is not NULL, this argument is ignored.

eval_fun_options

List of named arguments to pass to eval_fun.

vary_params

A vector of parameter names that are varied across in the Experiment.

curve

Either "ROC" or "PR" indicating whether to plot the ROC or Precision-Recall curve.

show

Character vector with elements being one of "boxplot", "point", "line", "bar", "errorbar", "ribbon" indicating what plot layer(s) to construct.

...

Not used.

Value

If interactive = TRUE, returns a plotly object if plot_by is NULL and a list of plotly objects if plot_by is not NULL. If interactive = FALSE, returns a ggplot object if plot_by is NULL and a list of ggplot objects if plot_by is not NULL.

See also

Other feature_selection_funs: eval_feature_importance_funs, eval_feature_selection_curve_funs, eval_feature_selection_err_funs, plot_feature_importance(), plot_feature_selection_err()

Examples

# generate example fit_results data
fit_results <- tibble::tibble(
  .rep = rep(1:2, times = 2),
  .dgp_name = c("DGP1", "DGP1", "DGP2", "DGP2"),
  .method_name = c("Method"),
  feature_info = lapply(
    1:4,
    FUN = function(i) {
      tibble::tibble(
        # feature names
        feature = c("featureA", "featureB", "featureC"),  
        # true feature support
        true_support = c(TRUE, FALSE, TRUE),  
        # estimated feature support
        est_support = c(TRUE, FALSE, FALSE),  
        # estimated feature importance scores
        est_importance = c(10, runif(2, min = -2, max = 2))  
      )
    }
  )
)

# generate example eval_results data
eval_results <- list(
  ROC = summarize_feature_selection_curve(
    fit_results, 
    curve = "ROC",
    nested_cols = "feature_info",
    truth_col = "true_support", 
    imp_col = "est_importance"
  ),
  PR = summarize_feature_selection_curve(
    fit_results, 
    curve = "PR",
    nested_cols = "feature_info",
    truth_col = "true_support", 
    imp_col = "est_importance"
  )
)

# create summary ROC/PR plots using pre-computed evaluation results
roc_plt <- plot_feature_selection_curve(eval_results = eval_results,
                                        eval_name = "ROC", curve = "ROC",
                                        show = c("line", "ribbon"))
pr_plt <- plot_feature_selection_curve(eval_results = eval_results,
                                       eval_name = "PR", curve = "PR",
                                       show = c("line", "ribbon"))
# or alternatively, create the same plots directly from fit results
roc_plt <- plot_feature_selection_curve(fit_results = fit_results,
                                        show = c("line", "ribbon"),
                                        curve = "ROC",
                                        eval_fun_options = list(
                                          nested_cols = "feature_info",
                                          truth_col = "true_support",
                                          imp_col = "est_importance"
                                        ))
pr_plt <- plot_feature_selection_curve(fit_results = fit_results,
                                       show = c("line", "ribbon"),
                                       curve = "PR",
                                       eval_fun_options = list(
                                         nested_cols = "feature_info",
                                         truth_col = "true_support",
                                         imp_col = "est_importance"
                                       ))

# can customize plot (see plot_eval_constructor() for possible arguments)
roc_plt <- plot_feature_selection_curve(eval_results = eval_results,
                                        eval_name = "ROC", curve = "ROC",
                                        show = c("line", "ribbon"),
                                        plot_by = ".dgp_name")