Note
Go to the end to download the full example code
Benchmark suites¶
How to list, download and upload benchmark suites.
If you want to learn more about benchmark suites, check out our brief introductory tutorial Benchmark suites or the OpenML benchmark docs.
# License: BSD 3-Clause
import uuid
import numpy as np
import openml
Listing suites¶
Use the output_format parameter to select output type
Default gives
dict, but we’ll usedataframeto obtain an easier-to-work-with data structure
suites = openml.study.list_suites(output_format="dataframe", status="all")
print(suites.head(n=10))
id alias ... creation_date creator
14 14 OpenML100 ... 2019-02-21 18:40:13 1
99 99 OpenML-CC18 ... 2019-02-21 18:47:13 1
218 218 AutoML-Benchmark ... 2019-05-02 13:35:08 869
219 219 FOREX ... 2019-06-04 00:45:17 1
225 225 OpenML-friendly ... 2019-09-16 19:41:46 1
236 236 a9ee1f0b2a4b48b6b6da1653fe92890e ... 2020-04-06 21:38:55 64
239 239 c638a5d3d31241179f9b4853951fdb79 ... 2020-04-19 22:15:30 64
240 240 e5e7f56c8655433eb2418c240ec8b8c0 ... 2020-04-28 02:30:38 2902
253 253 testecc18 ... 2020-09-01 00:57:54 8598
258 258 NaN ... 2020-09-30 08:30:00 8598
[10 rows x 7 columns]
Downloading suites¶
This is done based on the dataset ID.
suite = openml.study.get_suite(99)
print(suite)
OpenML Benchmark Suite
======================
ID..............: 99
Name............: OpenML-CC18 Curated Classification benchmark
Status..........: active
Main Entity Type: task
Study URL.......: https://www.openml.org/s/99
# of Data.......: 72
# of Tasks......: 72
Creator.........: https://www.openml.org/u/1
Upload Time.....: 2019-02-21 18:47:13
Suites also feature a description:
print(suite.description)
We advocate the use of curated, comprehensive benchmark suites of machine learning datasets, backed by standardized OpenML-based interfaces and complementary software toolkits written in Python, Java and R. We demonstrate how to easily execute comprehensive benchmarking studies using standardized OpenML-based benchmarking suites and complementary software toolkits written in Python, Java and R. Major distinguishing features of OpenML benchmark suites are (i) ease of use through standardized data formats, APIs, and existing client libraries; (ii) machine-readable meta-information regarding the contents of the suite; and (iii) online sharing of results, enabling large scale comparisons. As a first such suite, we propose the OpenML-CC18, a machine learning benchmark suite of 72 classification datasets carefully curated from the thousands of datasets on OpenML.
The inclusion criteria are:
* classification tasks on dense data set
independent observations
* number of classes larger or equal to 2, each class with at least 20 observations and ratio of minority to majority class must exceed 5%
* number of observations between 500 and 100000
* number of features after one-hot-encoding less than 5000
* no artificial data sets
* no subsets of larger data sets nor binarizations of other data sets
* no data sets which are perfectly predictable by using a single feature or by using a simple decision tree
* source or reference available
If you use this benchmarking suite, please cite:
Bernd Bischl, Giuseppe Casalicchio, Matthias Feurer, Frank Hutter, Michel Lang, Rafael G. Mantovani, Jan N. van Rijn and Joaquin Vanschoren. “OpenML Benchmarking Suites” arXiv:1708.03731v2 [stats.ML] (2019).
```
@article{oml-benchmarking-suites,
title={OpenML Benchmarking Suites},
author={Bernd Bischl and Giuseppe Casalicchio and Matthias Feurer and Frank Hutter and Michel Lang and Rafael G. Mantovani and Jan N. van Rijn and Joaquin Vanschoren},
year={2019},
journal={arXiv:1708.03731v2 [stat.ML]}
}
```
Suites are a container for tasks:
print(suite.tasks)
[3, 6, 11, 12, 14, 15, 16, 18, 22, 23, 28, 29, 31, 32, 37, 43, 45, 49, 53, 219, 2074, 2079, 3021, 3022, 3481, 3549, 3560, 3573, 3902, 3903, 3904, 3913, 3917, 3918, 7592, 9910, 9946, 9952, 9957, 9960, 9964, 9971, 9976, 9977, 9978, 9981, 9985, 10093, 10101, 14952, 14954, 14965, 14969, 14970, 125920, 125922, 146195, 146800, 146817, 146819, 146820, 146821, 146822, 146824, 146825, 167119, 167120, 167121, 167124, 167125, 167140, 167141]
And we can use the task listing functionality to learn more about them:
tasks = openml.tasks.list_tasks(output_format="dataframe")
# Using ``@`` in `pd.DataFrame.query <
# https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.query.html>`_
# accesses variables outside of the current dataframe.
tasks = tasks.query("tid in @suite.tasks")
print(tasks.describe().transpose())
count mean ... 75% max
tid 72.0 41849.305556 ... 42707.50 167141.0
did 72.0 10472.263889 ... 10594.25 41027.0
MajorityClassSize 72.0 4820.166667 ... 2796.00 48658.0
MaxNominalAttDistinctValues 71.0 8.718310 ... 10.00 71.0
MinorityClassSize 72.0 2199.777778 ... 1326.50 47662.0
NumberOfClasses 72.0 5.680556 ... 8.25 46.0
NumberOfFeatures 72.0 196.597222 ... 71.50 3073.0
NumberOfInstances 72.0 12148.972222 ... 6771.75 96320.0
NumberOfInstancesWithMissingValues 72.0 121.347222 ... 0.00 3772.0
NumberOfMissingValues 72.0 226.597222 ... 0.00 6465.0
NumberOfNumericFeatures 72.0 166.875000 ... 64.00 3072.0
NumberOfSymbolicFeatures 72.0 29.722222 ... 5.25 1556.0
target_feature_left 0.0 NaN ... NaN NaN
[13 rows x 8 columns]
We’ll use the test server for the rest of this tutorial.
Warning
This example uploads data. For that reason, this example connects to the test server at test.openml.org. This prevents the main server from crowding with example datasets, tasks, runs, and so on. The use of this test server can affect behaviour and performance of the OpenML-Python API.
openml.config.start_using_configuration_for_example()
/home/runner/work/openml-python/openml-python/openml/config.py:184: UserWarning: Switching to the test server https://test.openml.org/api/v1/xml to not upload results to the live server. Using the test server may result in reduced performance of the API!
warnings.warn(
Uploading suites¶
Uploading suites is as simple as uploading any kind of other OpenML entity - the only reason why we need so much code in this example is because we upload some random data.
# We'll take a random subset of at least ten tasks of all available tasks on
# the test server:
all_tasks = list(openml.tasks.list_tasks(output_format="dataframe")["tid"])
task_ids_for_suite = sorted(np.random.choice(all_tasks, replace=False, size=20))
# The study needs a machine-readable and unique alias. To obtain this,
# we simply generate a random uuid.
alias = uuid.uuid4().hex
new_suite = openml.study.create_benchmark_suite(
name="Test-Suite",
description="Test suite for the Python tutorial on benchmark suites",
task_ids=task_ids_for_suite,
alias=alias,
)
new_suite.publish()
print(new_suite)
/home/runner/work/openml-python/openml-python/openml/tasks/functions.py:112: RuntimeWarning: Could not create task type id for 10 due to error 10 is not a valid TaskType
warnings.warn(
/home/runner/work/openml-python/openml-python/openml/tasks/functions.py:112: RuntimeWarning: Could not create task type id for 11 due to error 11 is not a valid TaskType
warnings.warn(
/home/runner/work/openml-python/openml-python/openml/tasks/functions.py:263: RuntimeWarning: Could not create task type id for 11 due to error 11 is not a valid TaskType
warnings.warn(
/home/runner/work/openml-python/openml-python/openml/tasks/functions.py:263: RuntimeWarning: Could not create task type id for 10 due to error 10 is not a valid TaskType
warnings.warn(
OpenML Benchmark Suite
======================
ID..............: 782
Name............: Test-Suite
Status..........: None
Main Entity Type: task
Study URL.......: https://test.openml.org/s/782
# of Tasks......: 20
openml.config.stop_using_configuration_for_example()
Total running time of the script: ( 0 minutes 48.415 seconds)