dscience.ml.decision_frame

Module Contents

class dscience.ml.decision_frame.DecisionFrame

Bases: OrganizingFrame

An n × m matrix of probabilities (or scores) from a classifier. The n rows are samples, and the m columns are predictions. The values are the confidence or pobability of prediction. The single index column is named ‘correct_label’, and the single column name is named ‘label’. Practically, this is a Pandas wrapper around a scikit-learn decision_function that also has the predicted and correct class labels.

classmethod required_index_names(cls)
classmethod of(cls, correct_labels: Sequence[str], labels: Sequence[str], decision_function: np.array, sample_ids: Sequence[Any])

Wraps a decision function numpy array into a DecisionFrame instance complete with labels as names and columns. :param correct_labels: A length-n list of the correct labels for each of the n samples :param labels: A length-m list of class labels matching the predictions (columns) on probabilities :param decision_function: An n × m matrix of probabilities (or scores) from the classifier.

The rows are samples, and the columns are predictions. scikit-learn decision_functions (ex model.predict_proba) will output this.
Parameters:sample_ids – IDs (or names) of training examples for later reference; should be unique
Returns:A DecisionFrame
confusion(self)
accuracy(self)
classmethod read_csv(cls, path: PathLike, *args, **kwargs)
to_csv(self, path: PathLike, *args, **kwargs)