pdpbox.pdp.PDPIsolate

class pdpbox.pdp.PDPIsolate(n_classes, which_class, feature, feature_type, feature_grids, percentile_info, display_columns, ice_lines, pdp, count_data, hist_data)

Save pdp_isolate results

Parameters:
n_classes: integer or None

number of classes for classifier, None when it is a regressor

which_class: integer or None

for multi-class classifier, indicate which class the result belongs to

feature: string or list

which feature is calculated on, list for one-hot encoding features

feature_type: string

type of the feature

feature_grids: list

feature grids

percentile_info: list

percentile information for feature grids

display_columns: list

columns to display as xticklabels

ice_lines: pandas DataFrame

ICE lines

pdp: 1-d numpy array

calculated PDP values

count_data: pandas DataFrame

data points distribution

hist_data: 1-d numpy array

data points distribution for numeric features