docs: restoring Auto MPG example but using data stored in the repo

README.md

@@ -64,10 +64,11 @@ ci = fci.random_forest_error(
 
 ## Examples
 
-The examples (gallery below) demonstrates the package functionality with random forest classifiers and regression models.
-The regression examples use scikit-learn's bundled diabetes dataset, while the
-classifier example simulates how to add measurements of uncertainty to tasks
-like predicting spam emails. Keeping the regression data bundled makes
+The examples (gallery below) demonstrate the package functionality with random forest classifiers and regression models.
+The regression examples use a bundled copy of the
+[UCI Auto MPG dataset](https://doi.org/10.24432/C5859H), while the classifier
+example simulates how to add measurements of uncertainty to tasks like
+predicting spam emails. Keeping the regression data in the repository makes
 documentation builds reproducible without relying on an external data service.
 
 [Examples gallery](http://contrib.scikit-learn.org/forest-confidence-interval/auto_examples/index.html)

examples/README.txt

@@ -3,10 +3,13 @@
 Examples
 =========
 
-The examples use data from standard machine learning libraries to demonstrate
-how `forestci` can be used to calculate error bars on
+The examples use standard machine learning datasets to demonstrate how
+`forestci` can be used to calculate error bars on
 :class:`RandomForestRegressor` and :class:`RandomForestClassifier` objects. The
-regression example uses a data-set from the `UC Irvine Machine Learning Repository <https://archive.ics.uci.edu/ml/datasets/Auto+MPG>`_ with features of
-different cars and their MPG. The classification example generates synthetic
-data to simulate a task like that of a spam filter: classifying items into one
-of two categories (e.g., spam/non-spam) based on a number of features.
+regression examples use a bundled copy of the `Auto MPG dataset
+<https://doi.org/10.24432/C5859H>`_ from the UC Irvine Machine Learning
+Repository, with features of different cars and their MPG. The classification
+example generates synthetic data to simulate a task like that of a spam
+filter: classifying items into one of two categories (e.g., spam/non-spam)
+based on a number of features. The dataset attribution and license are
+documented in ``data/README.md``.

examples/data/README.md

@@ -0,0 +1,21 @@
+# Auto MPG dataset
+
+`auto_mpg.csv` is a copy of the Auto MPG dataset created by R. Quinlan
+(1993), obtained from [OpenML dataset 196](https://www.openml.org/d/196).
+The canonical UCI Machine Learning Repository record is:
+
+> Quinlan, R. (1993). *Auto MPG* [Dataset]. UCI Machine Learning Repository.
+> https://doi.org/10.24432/C5859H
+
+## Copyright and license
+
+The dataset is licensed under the
+[Creative Commons Attribution 4.0 International license (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/).
+The SPDX license identifier is `CC-BY-4.0`. The UCI record does not provide a
+separate copyright notice; copyright remains with the original rights
+holder(s).
+
+The OpenML representation omits the original car-name identifier. For inclusion
+here, the ARFF data were converted to CSV, and the `model` and `class` column
+names were changed to `model_year` and `mpg`, respectively. Missing horsepower
+values are represented by empty CSV fields.