Code

Submodules

pyBreakDown.explanation module

class pyBreakDown.explanation.AttrInfo(name, value, contribution, cumulative)
class pyBreakDown.explanation.ExplainerDirection[source]

An enumeration.

class pyBreakDown.explanation.Explanation(variable_names, variable_values, contributions, direction)[source]

Contains algorithm results, including contribiutions of each individual features.

text(fwidth=25, contwidth=20, cumulwidth=20, digits=2)[source]

Get user-friendly text from of explanation

fwidth : int
Width of column with feature names, in digits.
contwidth : int
Width of column with contributions, in digits.
cumulwidth : int
Width of column with cumulative values, in digits.
digits : int
Number of decimal places for values.
visualize(figsize=(7, 6), filename=None, dpi=90, fontsize=14)[source]

Get user friendly visualization of explanation

figsize : tuple int
Pyplot figure size
filename : string
Name of file to save the visualization. If not specified, standard pyplot.show() will be performed.
dpi : int
Digits per inch for saving to the file

pyBreakDown.explainer module

class pyBreakDown.explainer.Explainer(clf, data, colnames)[source]

Explainer object.

clf : np.array
Sklearn predicition model (regression or classification).
data : np.array
Baseline dataset for algorithm.
colnames : np.array
Dataset feature names.
explain(observation, direction, useIntercept=False, baseline=0)[source]

Make explanation for given observation and dataset.

Method works with any sklearn prediction model

observation : np.array
Observation to explain.
direction : str
Could be “up” or “down”. Decides the direction of algorithm.
useIntercept : bool
If set, baseline argument will be ignored and baseline will be set to intercept.
baseline : float
Baseline of explanation.
Explanation
Object that contains influences and descriptions of each relevant attribute.