Error in predict_profile when a Dataframe with MultiIndex is Used

[CahidArda]

I am calling predict_profile method with a dataframe which has multiple columns as index. This results in an error.

How to Replicate

This occurs when pandas version is after 1.3.0. It doesn't occur in 1.2.5 and before.

import numpy as np
data  = np.array([[242,902,3,435],
                  [125,684,3,143],
                  [162,284,3,124],
                  [712,844,3,145],
                  [122,864,3,114],
                  [155,100,3,25]])
target = np.array([723,554,932,543,654,345])

import pandas as pd
data = pd.DataFrame(data, columns=[f"col{i}" for i in range(4)])
data = data.set_index(["col1", "col2"])

from sklearn.tree import DecisionTreeClassifier
clf = DecisionTreeClassifier()
clf.fit(data, target)

import dalex as dx
exp = dx.Explainer(clf, data, target)
pred_profile = exp.predict_profile(data, y=target)

from dalex.model_explanations import AggregatedProfiles
aggregated_profile = AggregatedProfiles()
aggregated_profile.fit(pred_profile)
aggregated_profile.plot(show=True)

Error Message

Calculating ceteris paribus:   0%|                                                            | 0/2 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "file.py", line 20, in <module>
    exp.predict_profile(data, y=target)
  File "C:\Users\user\Documents\GitHub\DALEX\python\dalex\dalex\_explainer\object.py", line 404, in predict_profile
    _predict_profile.fit(self, new_observation, y, verbose)
  File "C:\Users\user\Documents\GitHub\DALEX\python\dalex\dalex\predict_explanations\_ceteris_paribus\object.py", line 126, in fit
    self.result, self.new_observation = utils.calculate_ceteris_paribus(
  File "C:\Users\user\Documents\GitHub\DALEX\python\dalex\dalex\predict_explanations\_ceteris_paribus\utils.py", line 23, in calculate_ceteris_paribus
    profiles = calculate_variable_profile(explainer.predict_function,
  File "C:\Users\user\Documents\GitHub\DALEX\python\dalex\dalex\predict_explanations\_ceteris_paribus\utils.py", line 66, in calculate_variable_profile
    profile.append(single_variable_profile(predict_function, model, data, variable, split_points))
  File "C:\Users\user\Documents\GitHub\DALEX\python\dalex\dalex\predict_explanations\_ceteris_paribus\utils.py", line 112, in single_variable_profile
    new_data.loc[:, '_ids_'] = ids
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\indexing.py", line 723, in __setitem__
    iloc._setitem_with_indexer(indexer, value, self.name)
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\indexing.py", line 1667, in _setitem_with_indexer
    self.obj[key] = value
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\frame.py", line 3607, in __setitem__
    self._set_item(key, value)
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\frame.py", line 3779, in _set_item
    value = self._sanitize_column(value)
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\frame.py", line 4505, in _sanitize_column
    return sanitize_array(value, self.index, copy=True, allow_2d=True)
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\construction.py", line 500, in sanitize_array
    data = extract_array(data, extract_numpy=True)
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\construction.py", line 423, in extract_array
    obj = obj.array
  File "C:\Users\user\anaconda3\envs\dalex\lib\site-packages\pandas\core\indexes\multi.py", line 725, in array
    raise ValueError(
ValueError: MultiIndex has no single backing array. Use 'MultiIndex.to_numpy()' to get a NumPy array of tuples.

Solution

I have found the same error in the issues of another project and they suggest downgrading pandas to 1.2.4. Downgrading pandas to 1.2.4 solved the issue for me too.

I noticed another thing while looking at the single_variable_profile function in predict_explanations._ceteris_paribus.utils file. Changing the following line from:

ids = np.repeat(data.index, split_points.shape[0])

to:

ids = np.repeat(data.index.values, split_points.shape[0])

solves the issue in pandas versions 1.2.5 and 1.3.0.

[hbaniecki]

Hi @CahidArda, would you like to make a pull request with the solution?

[CahidArda]

Hello @hbaniecki, I have created a PR for the change.