.. DO NOT EDIT. .. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY. .. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE: .. "tutorials/interpretability.py" .. LINE NUMBERS ARE GIVEN BELOW. .. only:: html .. note:: :class: sphx-glr-download-link-note :ref:`Go to the end ` to download the full example code. .. rst-class:: sphx-glr-example-title .. _sphx_glr_tutorials_interpretability.py: Model interpretability with TabICL ============================ TabICL comes with a fast approximations of SHAP values. It is much faster than using black-box shape routines on TabICL which is slow. Here we demo it on dataset on wages .. GENERATED FROM PYTHON SOURCE LINES 12-14 The dataset: wages --------------------- .. GENERATED FROM PYTHON SOURCE LINES 14-21 .. code-block:: Python from sklearn.datasets import fetch_openml survey = fetch_openml(data_id=534, as_frame=True) X = survey.data[survey.feature_names] .. GENERATED FROM PYTHON SOURCE LINES 22-23 A quick glance at the data with skrub's TableReport .. GENERATED FROM PYTHON SOURCE LINES 23-27 .. code-block:: Python import skrub skrub.TableReport(X) .. raw:: html

	EDUCATION	SOUTH	SEX	EXPERIENCE	UNION	AGE	RACE	OCCUPATION	SECTOR	MARR
0	8	no	female	21	not_member	35	Hispanic	Other	Manufacturing	Married
1	9	no	female	42	not_member	57	White	Other	Manufacturing	Married
2	12	no	male	1	not_member	19	White	Other	Manufacturing	Unmarried
3	12	no	male	4	not_member	22	White	Other	Other	Unmarried
4	12	no	male	17	not_member	35	White	Other	Other	Married

529	18	no	male	5	not_member	29	White	Professional	Other	Unmarried
530	12	no	female	33	not_member	51	Other	Professional	Other	Married
531	17	no	female	25	member	48	Other	Professional	Other	Married
532	12	yes	male	13	member	31	White	Professional	Other	Married
533	16	no	male	33	not_member	55	White	Professional	Manufacturing	Married

Column	Column name	dtype	Is sorted	Unique values	Mean	Std	Min	Median	Max
0	EDUCATION	Int64DType	False	17 (3.2%)	13.0	2.62	2	12	18
1	SOUTH	CategoricalDtype	False	2 (0.4%)
2	SEX	CategoricalDtype	False	2 (0.4%)
3	EXPERIENCE	Int64DType	False	52 (9.7%)	17.8	12.4	0	15	55
4	UNION	CategoricalDtype	False	2 (0.4%)
5	AGE	Int64DType	False	47 (8.8%)	36.8	11.7	18	35	64
6	RACE	CategoricalDtype	False	3 (0.6%)
7	OCCUPATION	CategoricalDtype	False	6 (1.1%)
8	SECTOR	CategoricalDtype	False	3 (0.6%)
9	MARR	CategoricalDtype	False	2 (0.4%)

Column 1	Column 2	Cramér's V	Pearson's Correlation
EXPERIENCE	AGE	0.637	0.978
SEX	OCCUPATION	0.418
OCCUPATION	SECTOR	0.368
EXPERIENCE	MARR	0.359
AGE	MARR	0.352
EDUCATION	OCCUPATION	0.319
UNION	OCCUPATION	0.240
EDUCATION	EXPERIENCE	0.199	-0.353
EDUCATION	SECTOR	0.191
SEX	SECTOR	0.182
EDUCATION	AGE	0.180	-0.150
SOUTH	EXPERIENCE	0.170
EDUCATION	RACE	0.167
EDUCATION	SOUTH	0.165
SEX	UNION	0.157
EXPERIENCE	UNION	0.156
SOUTH	AGE	0.155
EXPERIENCE	RACE	0.154
EDUCATION	SEX	0.148
EXPERIENCE	SECTOR	0.146

Please enable javascript

The skrub table reports need javascript to display correctly. If you are displaying a report in a Jupyter notebook and you see this message, you may need to re-execute the cell or to trust the notebook (button on the top right or "File > Trust notebook").

.. GENERATED FROM PYTHON SOURCE LINES 28-29 We need to convert the categorical features to numeric ones. We can do this with pandas' get_dummies .. GENERATED FROM PYTHON SOURCE LINES 29-33 .. code-block:: Python import pandas as pd X = pd.get_dummies(X, drop_first=True) .. GENERATED FROM PYTHON SOURCE LINES 34-35 The values to predict: wages .. GENERATED FROM PYTHON SOURCE LINES 35-37 .. code-block:: Python y = survey.target.values.ravel() .. GENERATED FROM PYTHON SOURCE LINES 38-39 Split out a test set .. GENERATED FROM PYTHON SOURCE LINES 39-43 .. code-block:: Python from sklearn.model_selection import train_test_split X_train, X_test, y_train, y_test = train_test_split(X, y) .. GENERATED FROM PYTHON SOURCE LINES 44-47 Our TabICL model ------------------ .. GENERATED FROM PYTHON SOURCE LINES 47-52 .. code-block:: Python from tabicl import TabICLRegressor clf = TabICLRegressor(n_estimators=4, device="cpu") clf.fit(X_train, y_train) .. rst-class:: sphx-glr-script-out .. code-block:: none Checkpoint 'tabicl-regressor-v2-20260212.ckpt' not cached. Downloading from Hugging Face Hub (jingang/TabICL). Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads. .. raw:: html

TabICLRegressor(device='cpu', n_estimators=4)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

TabICLRegressor

iFitted

Parameters

	n_estimators	4
	norm_methods	None
	feat_shuffle_method	'latin'
	outlier_threshold	4.0
	batch_size	8
	kv_cache	False
	model_path	None
	allow_auto_download	True
	checkpoint_version	'tabicl-regressor-v2-20260212.ckpt'
	device	'cpu'
	use_amp	'auto'
	use_fa3	'auto'
	offload_mode	'auto'
	disk_offload_dir	None
	random_state	42
	n_jobs	None
	verbose	False
	inference_config	None

.. GENERATED FROM PYTHON SOURCE LINES 53-60 Shap-like interpretability --------------------------- Use TabICL's fast approximations of shap-like values and plot them This part of the example requires to install the shap extra: pip install 'tabicl[shap] .. GENERATED FROM PYTHON SOURCE LINES 60-66 .. code-block:: Python from tabicl.shap import get_shap_values, plot_shap # Compute the shap values sv = get_shap_values(clf, X_test[:10]) .. rst-class:: sphx-glr-script-out .. code-block:: none PermutationExplainer explainer: 10%|█ | 1/10 [00:00` .. container:: sphx-glr-download sphx-glr-download-python :download:`Download Python source code: interpretability.py ` .. container:: sphx-glr-download sphx-glr-download-zip :download:`Download zipped: interpretability.zip ` .. only:: html .. rst-class:: sphx-glr-signature `Gallery generated by Sphinx-Gallery `_

EDUCATION

SOUTH

SEX

EXPERIENCE

UNION

AGE

RACE

OCCUPATION

SECTOR

MARR