ClassifierEvaluator#

ClassifierEvaluator#

class sklearn_evaluation.ClassifierEvaluator(estimator=None, y_true=None, y_pred=None, y_score=None, feature_names=None, target_names=None, estimator_name=None, X=None)#

Encapsulates results from an estimator on a testing set to provide a simplified API from other modules. All parameters are optional, just fill the ones you need for your analysis.

Parameters:
  • estimator (sklearn estimator) ā€“ Must have a feature_importances_ attribute.

  • y_true (array-like) ā€“ Target predicted classes (estimator predictions).

  • y_pred (array-like) ā€“ Correct target values (ground truth).

  • y_score (array-like) ā€“ Target scores (estimator predictions).

  • feature_names (array-like) ā€“ Feature names.

  • target_names (list) ā€“ List containing the names of the target classes

  • estimator_name (str) ā€“ Identifier for the model. This can be later used to identify the estimator when generating reports.

confusion_matrix()#

Confusion matrix plot

property estimator_class#

Estimator class (e.g. sklearn.ensemble.RandomForestClassifier)

property estimator_type#

Estimator name (e.g. RandomForestClassifier)

feature_importances()#

Feature importances plot

feature_importances_table()#

Feature importances table

html_serializable()#

Returns a EvaluatorHTMLSerializer instance, which is an object with the same methods and properties than a ClassifierEvaluator, but it returns HTML serialized versions of each (i.e. evaluator.feature_importances_table() returns a string with the table in HTML format, evaluator.confusion_matrix() returns a HTML image element with the image content encoded in base64), useful for generating reports using some template system

make_report(template=None)#

Make HTML report

Parameters:
  • template (str, or pathlib.Path, optional) ā€“ HTML or Markdown template with jinja2 format. If a pathlib.Path object is passed, the content of the file is read. Within the template, the evaluator is passed as ā€œeā€, so you can use things like {{e.confusion_matrix()}} or any other attribute/method. If None, a default template is used

  • style (str) ā€“ Path to a css file to apply style to the report. If None, no style will be applied

Returns:

Returns the contents of the report if path is None.

Return type:

Report

precision_at_proportions()#

Precision at proportions plot

precision_recall()#

Precision-recall plot

roc()#

ROC plot