fetch_openml_datasets

imbens.datasets.fetch_openml_datasets(target_type='all', imalance_type='all', cat_preprocess='onehot', data_home=None)

Fetches multiple datasets from OpenML [1] based on the specified criteria and preprocesses them.

Added in version 0.3.0.

Parameters:

target_type{‘all’, ‘binary’, ‘multiclass’}, optional (default=’all’)

The type of target variable:

• all: No filter on target variable type.

• binary: Datasets with binary target variables.

• multiclass: Datasets with multiclass target variables.

imalance_type{‘all’, ‘low’, ‘medium’, ‘high’, ‘extreme’}, optional (default=’all’)

The imbalance type of the dataset:

• all: No filter on imbalance level.

• low: Datasets with low imbalance ratio (less than 5:1).

• medium: Datasets with medium imbalance ratio (between 5:1 and 10:1).

• high: Datasets with high imbalance ratio (between 10:1 and 50:1).

• extreme: Datasets with extreme imbalance ratio (greater than 50:1).

cat_preprocess{‘drop’, ‘onehot’, ‘ordinal’}, optional (default=’onehot’)

The method for preprocessing categorical features:

• drop: Drop categorical features.

• onehot: One-hot encode categorical features.

• ordinal: Ordinal encode categorical features.

data_homestr or Path or None, optional (default=None)

The directory where the datasets should be stored. If None, it will use the default user cache directory.

Returns:

datasetsOrderedDict of Bunch object,

The ordered is ranked by the imbalance ratio of the dataset. Each Bunch object — referred as dataset — have the following attributes:

dataset.datandarray of shape (n_samples, n_features)

dataset.targetndarray of shape (n_samples,)

dataset.openml_idint (OpenML dataset ID)

dataset.IRfloat (imbalance ratio of the dataset)

Notes

This function fetches datasets based on target and imbalance type filters, preprocesses them, and caches the datasets locally. The characteristics of the available datasets are presented in the table below.

ID	Name	OpenML ID	Ratio	#S	#F
1	bwin_amlb	45717	2.01:1	530	13
2	mozilla4	1046	2.04:1	15,545	5
3	mc2	1054	2.10:1	161	39
4	wholesale-customers	1511	2.10:1	440	8
5	vertebra-column	1524	2.10:1	310	6
6	law-school-admission-bianry	43904	2.11:1	20,800	14
7	bank32nh	833	2.22:1	8,192	32
8	elevators	846	2.24:1	16,599	18
9	cpu_small	735	2.31:1	8,192	12
10	Credit_Approval_Classification	46503	2.33:1	1,000	50
11	house_8L	843	2.38:1	22,784	8
12	house_16H	821	2.38:1	22,784	16
13	phoneme	1489	2.41:1	5,404	5
14	ilpd-numeric	41945	2.49:1	583	10
15	planning-relax	1490	2.50:1	182	12
16	MiniBooNE	41150	2.56:1	130,064	50
17	machine_cpu	733	2.73:1	209	6
18	telco-customer-churn	42178	2.77:1	7,043	39
19	haberman	43	2.78:1	306	3
20	vehicle	994	2.88:1	846	18
21	cpu	796	2.94:1	209	36
22	ada	41156	3.03:1	4,147	48
23	adult	45068	3.18:1	48,842	107
24	blood-transfusion-service-center	1464	3.20:1	748	4
25	default-of-credit-card-clients	42477	3.52:1	30,000	23
26	Customer_Churn_Classification	46362	3.74:1	175,028	24
27	SPECTF	1600	3.85:1	267	44
28	Medical-Appointment-No-Shows	43439	3.95:1	110,527	36
29	JapaneseVowels	976	5.17:1	9,961	14
30	ibm-employee-attrition	43893	5.20:1	1,470	53
31	first-order-theorem-proving	1475	5.26:1	6,118	51
32	user-knowledge	1508	5.38:1	403	5
33	online-shoppers-intention	45560	5.46:1	12,330	28
34	kc1	1067	5.47:1	2,109	21
35	thoracic-surgery	1506	5.71:1	470	16
36	UCI_churn	44232	5.90:1	3,333	18
37	arsenic-female-bladder	949	5.99:1	559	4
38	okcupid_stem	45067	6.83:1	26,677	117
39	ecoli	40671	7.15:1	327	7
40	pc4	1049	7.19:1	1,458	37
41	bank-marketing	1558	7.68:1	4,521	48
42	Diabetes-130-Hospitals_(Fairlearn)	43903	7.96:1	101,766	50
43	Otto-Group-Product-Classification-Challenge	45548	8.36:1	61,878	93
44	eucalyptus	43925	8.54:1	4,331	26
45	pendigits	1019	8.61:1	10,992	16
46	pc3	1050	8.77:1	1,563	37
47	page-blocks-bin	1021	8.77:1	5,473	10
48	optdigits	980	8.83:1	5,620	64
49	mfeat-karhunen	1020	9.00:1	2,000	64
50	mfeat-fourier	971	9.00:1	2,000	76
51	mfeat-zernike	995	9.00:1	2,000	47
52	Pulsar-Dataset-HTRU2	45558	9.92:1	17,898	8
53	vowel	1016	10.00:1	990	26
54	heart-h	1565	12.53:1	294	13
55	pc1	1068	13.40:1	1,109	21
56	seismic-bumps	45562	14.20:1	2,584	22
57	ozone-level-8hr	1487	14.84:1	2,534	72
58	microaggregation2	41671	15.02:1	20,000	20
59	Sick_numeric	41946	15.33:1	3,772	29
60	insurance_company	46281	15.76:1	9,822	85
61	wilt	40983	17.54:1	4,839	5
62	Click_prediction_small	1217	21.37:1	149,639	11
63	jannis	41168	22.83:1	83,733	54
64	letter	977	23.60:1	20,000	16
65	walking-activity	1509	24.14:1	149,332	4
66	helena	41169	36.08:1	65,196	27
67	mammography	310	42.01:1	11,183	6
68	dis	40713	64.03:1	3,772	29
69	Satellite	40900	67.00:1	5,100	36
70	Employee-Turnover-at-TECHCO	43551	68.74:1	34,452	9
71	page-blocks	30	175.46:1	5,473	10
72	allbp	40707	257.79:1	3,772	29
73	CreditCardFraudDetection	42397	577.88:1	284,807	30

References

[1]

Vanschoren, J., Van Rijn, J. N., Bischl, B., & Torgo, L. (2014). OpenML: networked science in machine learning. ACM SIGKDD Explorations Newsletter, 15(2), 49-60.