-
Notifications
You must be signed in to change notification settings - Fork 81
/
Copy pathensemble.html
2475 lines (2218 loc) · 234 KB
/
ensemble.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html lang="en" data-content_root="../" >
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" /><meta name="viewport" content="width=device-width, initial-scale=1" />
<meta property="og:title" content="1.11. Ensembles: Gradient boosting, random forests, bagging, voting, stacking" />
<meta property="og:type" content="website" />
<meta property="og:url" content="https://scikit-learn/stable/modules/ensemble.html" />
<meta property="og:site_name" content="scikit-learn" />
<meta property="og:description" content="Ensemble methods combine the predictions of several base estimators built with a given learning algorithm in order to improve generalizability / robustness over a single estimator. Two very famous ..." />
<meta property="og:image" content="https://scikit-learn/stable/_images/sphx_glr_plot_gradient_boosting_regression_001.png" />
<meta property="og:image:alt" content="scikit-learn" />
<meta name="description" content="Ensemble methods combine the predictions of several base estimators built with a given learning algorithm in order to improve generalizability / robustness over a single estimator. Two very famous ..." />
<title>1.11. Ensembles: Gradient boosting, random forests, bagging, voting, stacking — scikit-learn 1.7.dev0 documentation</title>
<script data-cfasync="false">
document.documentElement.dataset.mode = localStorage.getItem("mode") || "";
document.documentElement.dataset.theme = localStorage.getItem("theme") || "";
</script>
<!--
this give us a css class that will be invisible only if js is disabled
-->
<noscript>
<style>
.pst-js-only { display: none !important; }
</style>
</noscript>
<!-- Loaded before other Sphinx assets -->
<link href="../_static/styles/theme.css?digest=8878045cc6db502f8baf" rel="stylesheet" />
<link href="../_static/styles/pydata-sphinx-theme.css?digest=8878045cc6db502f8baf" rel="stylesheet" />
<link rel="stylesheet" type="text/css" href="../_static/pygments.css?v=8f2a1f02" />
<link rel="stylesheet" type="text/css" href="../_static/copybutton.css?v=76b2166b" />
<link rel="stylesheet" type="text/css" href="../_static/plot_directive.css" />
<link rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Vibur" />
<link rel="stylesheet" type="text/css" href="../_static/jupyterlite_sphinx.css?v=2c9f8f05" />
<link rel="stylesheet" type="text/css" href="../_static/sg_gallery.css?v=d2d258e8" />
<link rel="stylesheet" type="text/css" href="../_static/sg_gallery-binder.css?v=f4aeca0c" />
<link rel="stylesheet" type="text/css" href="../_static/sg_gallery-dataframe.css?v=2082cf3c" />
<link rel="stylesheet" type="text/css" href="../_static/sg_gallery-rendered-html.css?v=1277b6f3" />
<link rel="stylesheet" type="text/css" href="../_static/sphinx-design.min.css?v=95c83b7e" />
<link rel="stylesheet" type="text/css" href="../_static/styles/colors.css?v=cc94ab7d" />
<link rel="stylesheet" type="text/css" href="../_static/styles/custom.css?v=8f525996" />
<!-- So that users can add custom icons -->
<script src="../_static/scripts/fontawesome.js?digest=8878045cc6db502f8baf"></script>
<!-- Pre-loaded scripts that we'll load fully later -->
<link rel="preload" as="script" href="../_static/scripts/bootstrap.js?digest=8878045cc6db502f8baf" />
<link rel="preload" as="script" href="../_static/scripts/pydata-sphinx-theme.js?digest=8878045cc6db502f8baf" />
<script src="../_static/documentation_options.js?v=473747f4"></script>
<script src="../_static/doctools.js?v=9bcbadda"></script>
<script src="../_static/sphinx_highlight.js?v=dc90522c"></script>
<script src="../_static/clipboard.min.js?v=a7894cd8"></script>
<script src="../_static/copybutton.js?v=97f0b27d"></script>
<script src="../_static/jupyterlite_sphinx.js?v=96e329c5"></script>
<script src="../_static/design-tabs.js?v=f930bc37"></script>
<script data-domain="scikit-learn.org" defer="defer" src="https://views.scientific-python.org/js/script.js"></script>
<script async="async" src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-chtml.js"></script>
<script>DOCUMENTATION_OPTIONS.pagename = 'modules/ensemble';</script>
<script>
DOCUMENTATION_OPTIONS.theme_version = '0.16.1';
DOCUMENTATION_OPTIONS.theme_switcher_json_url = 'https://scikit-learn.org/dev/_static/versions.json';
DOCUMENTATION_OPTIONS.theme_switcher_version_match = '1.7.dev0';
DOCUMENTATION_OPTIONS.show_version_warning_banner =
true;
</script>
<script src="../_static/scripts/dropdown.js?v=d6825577"></script>
<script src="../_static/scripts/version-switcher.js?v=a6dd8357"></script>
<script src="../_static/scripts/sg_plotly_resize.js?v=2167d4db"></script>
<link rel="canonical" href="https://scikit-learn.org/stable/modules/ensemble.html" />
<link rel="icon" href="../_static/favicon.ico"/>
<link rel="author" title="About these documents" href="../about.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="1.12. Multiclass and multioutput algorithms" href="multiclass.html" />
<link rel="prev" title="1.10. Decision Trees" href="tree.html" />
<meta name="viewport" content="width=device-width, initial-scale=1"/>
<meta name="docsearch:language" content="en"/>
<meta name="docsearch:version" content="1.7" />
</head>
<body data-bs-spy="scroll" data-bs-target=".bd-toc-nav" data-offset="180" data-bs-root-margin="0px 0px -60%" data-default-mode="">
<div id="pst-skip-link" class="skip-link d-print-none"><a href="#main-content">Skip to main content</a></div>
<div id="pst-scroll-pixel-helper"></div>
<button type="button" class="btn rounded-pill" id="pst-back-to-top">
<i class="fa-solid fa-arrow-up"></i>Back to top</button>
<dialog id="pst-search-dialog">
<form class="bd-search d-flex align-items-center"
action="../search.html"
method="get">
<i class="fa-solid fa-magnifying-glass"></i>
<input type="search"
class="form-control"
name="q"
placeholder="Search the docs ..."
aria-label="Search the docs ..."
autocomplete="off"
autocorrect="off"
autocapitalize="off"
spellcheck="false"/>
<span class="search-button__kbd-shortcut"><kbd class="kbd-shortcut__modifier">Ctrl</kbd>+<kbd>K</kbd></span>
</form>
</dialog>
<div class="pst-async-banner-revealer d-none">
<aside id="bd-header-version-warning" class="d-none d-print-none" aria-label="Version warning"></aside>
</div>
<header class="bd-header navbar navbar-expand-lg bd-navbar d-print-none">
<div class="bd-header__inner bd-page-width">
<button class="pst-navbar-icon sidebar-toggle primary-toggle" aria-label="Site navigation">
<span class="fa-solid fa-bars"></span>
</button>
<div class=" navbar-header-items__start">
<div class="navbar-item">
<a class="navbar-brand logo" href="../index.html">
<img src="../_static/scikit-learn-logo-small.png" class="logo__image only-light" alt="scikit-learn homepage"/>
<img src="../_static/scikit-learn-logo-small.png" class="logo__image only-dark pst-js-only" alt="scikit-learn homepage"/>
</a></div>
</div>
<div class=" navbar-header-items">
<div class="me-auto navbar-header-items__center">
<div class="navbar-item">
<nav>
<ul class="bd-navbar-elements navbar-nav">
<li class="nav-item ">
<a class="nav-link nav-internal" href="../install.html">
Install
</a>
</li>
<li class="nav-item current active">
<a class="nav-link nav-internal" href="../user_guide.html">
User Guide
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../api/index.html">
API
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../auto_examples/index.html">
Examples
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-external" href="https://blog.scikit-learn.org/">
Community
</a>
</li>
<li class="nav-item dropdown">
<button class="btn dropdown-toggle nav-item" type="button"
data-bs-toggle="dropdown" aria-expanded="false"
aria-controls="pst-nav-more-links">
More
</button>
<ul id="pst-nav-more-links" class="dropdown-menu">
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../getting_started.html">
Getting Started
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../whats_new.html">
Release History
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../glossary.html">
Glossary
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../developers/index.html">
Development
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../faq.html">
FAQ
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../support.html">
Support
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../related_projects.html">
Related Projects
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../roadmap.html">
Roadmap
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../governance.html">
Governance
</a>
</li>
<li class=" ">
<a class="nav-link dropdown-item nav-internal" href="../about.html">
About us
</a>
</li>
</ul>
</li>
</ul>
</nav></div>
</div>
<div class="navbar-header-items__end">
<div class="navbar-item navbar-persistent--container">
<button class="btn btn-sm pst-navbar-icon search-button search-button__button pst-js-only" title="Search" aria-label="Search" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="fa-solid fa-magnifying-glass fa-lg"></i>
</button>
</div>
<div class="navbar-item">
<button class="btn btn-sm nav-link pst-navbar-icon theme-switch-button pst-js-only" aria-label="Color mode" data-bs-title="Color mode" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="theme-switch fa-solid fa-sun fa-lg" data-mode="light" title="Light"></i>
<i class="theme-switch fa-solid fa-moon fa-lg" data-mode="dark" title="Dark"></i>
<i class="theme-switch fa-solid fa-circle-half-stroke fa-lg" data-mode="auto" title="System Settings"></i>
</button></div>
<div class="navbar-item"><ul class="navbar-icon-links"
aria-label="Icon Links">
<li class="nav-item">
<a href="https://github.com/scikit-learn/scikit-learn" title="GitHub" class="nav-link pst-navbar-icon" rel="noopener" target="_blank" data-bs-toggle="tooltip" data-bs-placement="bottom"><i class="fa-brands fa-square-github fa-lg" aria-hidden="true"></i>
<span class="sr-only">GitHub</span></a>
</li>
</ul></div>
<div class="navbar-item">
<div class="version-switcher__container dropdown pst-js-only">
<button id="pst-version-switcher-button-2"
type="button"
class="version-switcher__button btn btn-sm dropdown-toggle"
data-bs-toggle="dropdown"
aria-haspopup="listbox"
aria-controls="pst-version-switcher-list-2"
aria-label="Version switcher list"
>
Choose version <!-- this text may get changed later by javascript -->
<span class="caret"></span>
</button>
<div id="pst-version-switcher-list-2"
class="version-switcher__menu dropdown-menu list-group-flush py-0"
role="listbox" aria-labelledby="pst-version-switcher-button-2">
<!-- dropdown will be populated by javascript on page load -->
</div>
</div></div>
</div>
</div>
<div class="navbar-persistent--mobile">
<button class="btn btn-sm pst-navbar-icon search-button search-button__button pst-js-only" title="Search" aria-label="Search" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="fa-solid fa-magnifying-glass fa-lg"></i>
</button>
</div>
<button class="pst-navbar-icon sidebar-toggle secondary-toggle" aria-label="On this page">
<span class="fa-solid fa-outdent"></span>
</button>
</div>
</header>
<div class="bd-container">
<div class="bd-container__inner bd-page-width">
<dialog id="pst-primary-sidebar-modal"></dialog>
<div id="pst-primary-sidebar" class="bd-sidebar-primary bd-sidebar">
<div class="sidebar-header-items sidebar-primary__section">
<div class="sidebar-header-items__center">
<div class="navbar-item">
<nav>
<ul class="bd-navbar-elements navbar-nav">
<li class="nav-item ">
<a class="nav-link nav-internal" href="../install.html">
Install
</a>
</li>
<li class="nav-item current active">
<a class="nav-link nav-internal" href="../user_guide.html">
User Guide
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../api/index.html">
API
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../auto_examples/index.html">
Examples
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-external" href="https://blog.scikit-learn.org/">
Community
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../getting_started.html">
Getting Started
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../whats_new.html">
Release History
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../glossary.html">
Glossary
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../developers/index.html">
Development
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../faq.html">
FAQ
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../support.html">
Support
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../related_projects.html">
Related Projects
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../roadmap.html">
Roadmap
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../governance.html">
Governance
</a>
</li>
<li class="nav-item ">
<a class="nav-link nav-internal" href="../about.html">
About us
</a>
</li>
</ul>
</nav></div>
</div>
<div class="sidebar-header-items__end">
<div class="navbar-item">
<button class="btn btn-sm nav-link pst-navbar-icon theme-switch-button pst-js-only" aria-label="Color mode" data-bs-title="Color mode" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="theme-switch fa-solid fa-sun fa-lg" data-mode="light" title="Light"></i>
<i class="theme-switch fa-solid fa-moon fa-lg" data-mode="dark" title="Dark"></i>
<i class="theme-switch fa-solid fa-circle-half-stroke fa-lg" data-mode="auto" title="System Settings"></i>
</button></div>
<div class="navbar-item"><ul class="navbar-icon-links"
aria-label="Icon Links">
<li class="nav-item">
<a href="https://github.com/scikit-learn/scikit-learn" title="GitHub" class="nav-link pst-navbar-icon" rel="noopener" target="_blank" data-bs-toggle="tooltip" data-bs-placement="bottom"><i class="fa-brands fa-square-github fa-lg" aria-hidden="true"></i>
<span class="sr-only">GitHub</span></a>
</li>
</ul></div>
<div class="navbar-item">
<div class="version-switcher__container dropdown pst-js-only">
<button id="pst-version-switcher-button-3"
type="button"
class="version-switcher__button btn btn-sm dropdown-toggle"
data-bs-toggle="dropdown"
aria-haspopup="listbox"
aria-controls="pst-version-switcher-list-3"
aria-label="Version switcher list"
>
Choose version <!-- this text may get changed later by javascript -->
<span class="caret"></span>
</button>
<div id="pst-version-switcher-list-3"
class="version-switcher__menu dropdown-menu list-group-flush py-0"
role="listbox" aria-labelledby="pst-version-switcher-button-3">
<!-- dropdown will be populated by javascript on page load -->
</div>
</div></div>
</div>
</div>
<div class="sidebar-primary-items__start sidebar-primary__section">
<div class="sidebar-primary-item">
<nav class="bd-docs-nav bd-links"
aria-label="Section Navigation">
<p class="bd-links__title" role="heading" aria-level="1">Section Navigation</p>
<div class="bd-toc-item navbar-nav"><ul class="current nav bd-sidenav">
<li class="toctree-l1 current active has-children"><a class="reference internal" href="../supervised_learning.html">1. Supervised learning</a><details open="open"><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul class="current">
<li class="toctree-l2"><a class="reference internal" href="linear_model.html">1.1. Linear Models</a></li>
<li class="toctree-l2"><a class="reference internal" href="lda_qda.html">1.2. Linear and Quadratic Discriminant Analysis</a></li>
<li class="toctree-l2"><a class="reference internal" href="kernel_ridge.html">1.3. Kernel ridge regression</a></li>
<li class="toctree-l2"><a class="reference internal" href="svm.html">1.4. Support Vector Machines</a></li>
<li class="toctree-l2"><a class="reference internal" href="sgd.html">1.5. Stochastic Gradient Descent</a></li>
<li class="toctree-l2"><a class="reference internal" href="neighbors.html">1.6. Nearest Neighbors</a></li>
<li class="toctree-l2"><a class="reference internal" href="gaussian_process.html">1.7. Gaussian Processes</a></li>
<li class="toctree-l2"><a class="reference internal" href="cross_decomposition.html">1.8. Cross decomposition</a></li>
<li class="toctree-l2"><a class="reference internal" href="naive_bayes.html">1.9. Naive Bayes</a></li>
<li class="toctree-l2"><a class="reference internal" href="tree.html">1.10. Decision Trees</a></li>
<li class="toctree-l2 current active"><a class="current reference internal" href="#">1.11. Ensembles: Gradient boosting, random forests, bagging, voting, stacking</a></li>
<li class="toctree-l2"><a class="reference internal" href="multiclass.html">1.12. Multiclass and multioutput algorithms</a></li>
<li class="toctree-l2"><a class="reference internal" href="feature_selection.html">1.13. Feature selection</a></li>
<li class="toctree-l2"><a class="reference internal" href="semi_supervised.html">1.14. Semi-supervised learning</a></li>
<li class="toctree-l2"><a class="reference internal" href="isotonic.html">1.15. Isotonic regression</a></li>
<li class="toctree-l2"><a class="reference internal" href="calibration.html">1.16. Probability calibration</a></li>
<li class="toctree-l2"><a class="reference internal" href="neural_networks_supervised.html">1.17. Neural network models (supervised)</a></li>
</ul>
</details></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../unsupervised_learning.html">2. Unsupervised learning</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="mixture.html">2.1. Gaussian mixture models</a></li>
<li class="toctree-l2"><a class="reference internal" href="manifold.html">2.2. Manifold learning</a></li>
<li class="toctree-l2"><a class="reference internal" href="clustering.html">2.3. Clustering</a></li>
<li class="toctree-l2"><a class="reference internal" href="biclustering.html">2.4. Biclustering</a></li>
<li class="toctree-l2"><a class="reference internal" href="decomposition.html">2.5. Decomposing signals in components (matrix factorization problems)</a></li>
<li class="toctree-l2"><a class="reference internal" href="covariance.html">2.6. Covariance estimation</a></li>
<li class="toctree-l2"><a class="reference internal" href="outlier_detection.html">2.7. Novelty and Outlier Detection</a></li>
<li class="toctree-l2"><a class="reference internal" href="density.html">2.8. Density Estimation</a></li>
<li class="toctree-l2"><a class="reference internal" href="neural_networks_unsupervised.html">2.9. Neural network models (unsupervised)</a></li>
</ul>
</details></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../model_selection.html">3. Model selection and evaluation</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="cross_validation.html">3.1. Cross-validation: evaluating estimator performance</a></li>
<li class="toctree-l2"><a class="reference internal" href="grid_search.html">3.2. Tuning the hyper-parameters of an estimator</a></li>
<li class="toctree-l2"><a class="reference internal" href="classification_threshold.html">3.3. Tuning the decision threshold for class prediction</a></li>
<li class="toctree-l2"><a class="reference internal" href="model_evaluation.html">3.4. Metrics and scoring: quantifying the quality of predictions</a></li>
<li class="toctree-l2"><a class="reference internal" href="learning_curve.html">3.5. Validation curves: plotting scores to evaluate models</a></li>
</ul>
</details></li>
<li class="toctree-l1"><a class="reference internal" href="../metadata_routing.html">4. Metadata Routing</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../inspection.html">5. Inspection</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="partial_dependence.html">5.1. Partial Dependence and Individual Conditional Expectation plots</a></li>
<li class="toctree-l2"><a class="reference internal" href="permutation_importance.html">5.2. Permutation feature importance</a></li>
</ul>
</details></li>
<li class="toctree-l1"><a class="reference internal" href="../visualizations.html">6. Visualizations</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../data_transforms.html">7. Dataset transformations</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="compose.html">7.1. Pipelines and composite estimators</a></li>
<li class="toctree-l2"><a class="reference internal" href="feature_extraction.html">7.2. Feature extraction</a></li>
<li class="toctree-l2"><a class="reference internal" href="preprocessing.html">7.3. Preprocessing data</a></li>
<li class="toctree-l2"><a class="reference internal" href="impute.html">7.4. Imputation of missing values</a></li>
<li class="toctree-l2"><a class="reference internal" href="unsupervised_reduction.html">7.5. Unsupervised dimensionality reduction</a></li>
<li class="toctree-l2"><a class="reference internal" href="random_projection.html">7.6. Random Projection</a></li>
<li class="toctree-l2"><a class="reference internal" href="kernel_approximation.html">7.7. Kernel Approximation</a></li>
<li class="toctree-l2"><a class="reference internal" href="metrics.html">7.8. Pairwise metrics, Affinities and Kernels</a></li>
<li class="toctree-l2"><a class="reference internal" href="preprocessing_targets.html">7.9. Transforming the prediction target (<code class="docutils literal notranslate"><span class="pre">y</span></code>)</a></li>
</ul>
</details></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../datasets.html">8. Dataset loading utilities</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../datasets/toy_dataset.html">8.1. Toy datasets</a></li>
<li class="toctree-l2"><a class="reference internal" href="../datasets/real_world.html">8.2. Real world datasets</a></li>
<li class="toctree-l2"><a class="reference internal" href="../datasets/sample_generators.html">8.3. Generated datasets</a></li>
<li class="toctree-l2"><a class="reference internal" href="../datasets/loading_other_datasets.html">8.4. Loading other datasets</a></li>
</ul>
</details></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../computing.html">9. Computing with scikit-learn</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../computing/scaling_strategies.html">9.1. Strategies to scale computationally: bigger data</a></li>
<li class="toctree-l2"><a class="reference internal" href="../computing/computational_performance.html">9.2. Computational Performance</a></li>
<li class="toctree-l2"><a class="reference internal" href="../computing/parallelism.html">9.3. Parallelism, resource management, and configuration</a></li>
</ul>
</details></li>
<li class="toctree-l1"><a class="reference internal" href="../model_persistence.html">10. Model persistence</a></li>
<li class="toctree-l1"><a class="reference internal" href="../common_pitfalls.html">11. Common pitfalls and recommended practices</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../dispatching.html">12. Dispatching</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="array_api.html">12.1. Array API support (experimental)</a></li>
</ul>
</details></li>
<li class="toctree-l1"><a class="reference internal" href="../machine_learning_map.html">13. Choosing the right estimator</a></li>
<li class="toctree-l1"><a class="reference internal" href="../presentations.html">14. External Resources, Videos and Talks</a></li>
</ul>
</div>
</nav></div>
</div>
<div class="sidebar-primary-items__end sidebar-primary__section">
</div>
</div>
<main id="main-content" class="bd-main" role="main">
<div class="bd-content">
<div class="bd-article-container">
<div class="bd-header-article d-print-none">
<div class="header-article-items header-article__inner">
<div class="header-article-items__start">
<div class="header-article-item">
<nav aria-label="Breadcrumb" class="d-print-none">
<ul class="bd-breadcrumbs">
<li class="breadcrumb-item breadcrumb-home">
<a href="../index.html" class="nav-link" aria-label="Home">
<i class="fa-solid fa-home"></i>
</a>
</li>
<li class="breadcrumb-item"><a href="../user_guide.html" class="nav-link">User Guide</a></li>
<li class="breadcrumb-item"><a href="../supervised_learning.html" class="nav-link"><span class="section-number">1. </span>Supervised learning</a></li>
<li class="breadcrumb-item active" aria-current="page"><span class="ellipsis"><span class="section-number">1.11. </span>Ensembles: Gradient boosting, random forests, bagging, voting, stacking</span></li>
</ul>
</nav>
</div>
</div>
</div>
</div>
<div id="searchbox"></div>
<article class="bd-article">
<section id="ensembles-gradient-boosting-random-forests-bagging-voting-stacking">
<span id="ensemble"></span><h1><span class="section-number">1.11. </span>Ensembles: Gradient boosting, random forests, bagging, voting, stacking<a class="headerlink" href="#ensembles-gradient-boosting-random-forests-bagging-voting-stacking" title="Link to this heading">#</a></h1>
<p><strong>Ensemble methods</strong> combine the predictions of several
base estimators built with a given learning algorithm in order to improve
generalizability / robustness over a single estimator.</p>
<p>Two very famous examples of ensemble methods are <a class="reference internal" href="#gradient-boosting"><span class="std std-ref">gradient-boosted trees</span></a> and <a class="reference internal" href="#forest"><span class="std std-ref">random forests</span></a>.</p>
<p>More generally, ensemble models can be applied to any base learner beyond
trees, in averaging methods such as <a class="reference internal" href="#bagging"><span class="std std-ref">Bagging methods</span></a>,
<a class="reference internal" href="#stacking"><span class="std std-ref">model stacking</span></a>, or <a class="reference internal" href="#voting-classifier"><span class="std std-ref">Voting</span></a>, or in
boosting, as <a class="reference internal" href="#adaboost"><span class="std std-ref">AdaBoost</span></a>.</p>
<section id="gradient-boosted-trees">
<span id="gradient-boosting"></span><h2><span class="section-number">1.11.1. </span>Gradient-boosted trees<a class="headerlink" href="#gradient-boosted-trees" title="Link to this heading">#</a></h2>
<p><a class="reference external" href="https://en.wikipedia.org/wiki/Gradient_boosting">Gradient Tree Boosting</a>
or Gradient Boosted Decision Trees (GBDT) is a generalization
of boosting to arbitrary differentiable loss functions, see the seminal work of
<a class="reference internal" href="#friedman2001" id="id1"><span>[Friedman2001]</span></a>. GBDT is an excellent model for both regression and
classification, in particular for tabular data.</p>
<aside class="topic">
<p class="topic-title"><a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingClassifier</span></code></a> vs <a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingClassifier.html#sklearn.ensemble.HistGradientBoostingClassifier" title="sklearn.ensemble.HistGradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingClassifier</span></code></a></p>
<p>Scikit-learn provides two implementations of gradient-boosted trees:
<a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingClassifier.html#sklearn.ensemble.HistGradientBoostingClassifier" title="sklearn.ensemble.HistGradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingClassifier</span></code></a> vs
<a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingClassifier</span></code></a> for classification, and the
corresponding classes for regression. The former can be <strong>orders of
magnitude faster</strong> than the latter when the number of samples is
larger than tens of thousands of samples.</p>
<p>Missing values and categorical data are natively supported by the
Hist… version, removing the need for additional preprocessing such as
imputation.</p>
<p><a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingClassifier</span></code></a> and
<a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor" title="sklearn.ensemble.GradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingRegressor</span></code></a> might be preferred for small sample
sizes since binning may lead to split points that are too approximate
in this setting.</p>
</aside>
<section id="histogram-based-gradient-boosting">
<span id="id2"></span><h3><span class="section-number">1.11.1.1. </span>Histogram-Based Gradient Boosting<a class="headerlink" href="#histogram-based-gradient-boosting" title="Link to this heading">#</a></h3>
<p>Scikit-learn 0.21 introduced two new implementations of
gradient boosted trees, namely <a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingClassifier.html#sklearn.ensemble.HistGradientBoostingClassifier" title="sklearn.ensemble.HistGradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingClassifier</span></code></a>
and <a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingRegressor.html#sklearn.ensemble.HistGradientBoostingRegressor" title="sklearn.ensemble.HistGradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingRegressor</span></code></a>, inspired by
<a class="reference external" href="https://github.com/Microsoft/LightGBM">LightGBM</a> (See <a class="reference internal" href="#lightgbm" id="id3"><span>[LightGBM]</span></a>).</p>
<p>These histogram-based estimators can be <strong>orders of magnitude faster</strong>
than <a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingClassifier</span></code></a> and
<a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor" title="sklearn.ensemble.GradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingRegressor</span></code></a> when the number of samples is larger
than tens of thousands of samples.</p>
<p>They also have built-in support for missing values, which avoids the need
for an imputer.</p>
<p>These fast estimators first bin the input samples <code class="docutils literal notranslate"><span class="pre">X</span></code> into
integer-valued bins (typically 256 bins) which tremendously reduces the
number of splitting points to consider, and allows the algorithm to
leverage integer-based data structures (histograms) instead of relying on
sorted continuous values when building the trees. The API of these
estimators is slightly different, and some of the features from
<a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingClassifier</span></code></a> and <a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor" title="sklearn.ensemble.GradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingRegressor</span></code></a>
are not yet supported, for instance some loss functions.</p>
<p class="rubric">Examples</p>
<ul class="simple">
<li><p><a class="reference internal" href="../auto_examples/inspection/plot_partial_dependence.html#sphx-glr-auto-examples-inspection-plot-partial-dependence-py"><span class="std std-ref">Partial Dependence and Individual Conditional Expectation Plots</span></a></p></li>
<li><p><a class="reference internal" href="../auto_examples/ensemble/plot_forest_hist_grad_boosting_comparison.html#sphx-glr-auto-examples-ensemble-plot-forest-hist-grad-boosting-comparison-py"><span class="std std-ref">Comparing Random Forests and Histogram Gradient Boosting models</span></a></p></li>
</ul>
<section id="usage">
<h4><span class="section-number">1.11.1.1.1. </span>Usage<a class="headerlink" href="#usage" title="Link to this heading">#</a></h4>
<p>Most of the parameters are unchanged from
<a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" title="sklearn.ensemble.GradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingClassifier</span></code></a> and <a class="reference internal" href="generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor" title="sklearn.ensemble.GradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">GradientBoostingRegressor</span></code></a>.
One exception is the <code class="docutils literal notranslate"><span class="pre">max_iter</span></code> parameter that replaces <code class="docutils literal notranslate"><span class="pre">n_estimators</span></code>, and
controls the number of iterations of the boosting process:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="kn">from</span><span class="w"> </span><span class="nn">sklearn.ensemble</span><span class="w"> </span><span class="kn">import</span> <span class="n">HistGradientBoostingClassifier</span>
<span class="gp">>>> </span><span class="kn">from</span><span class="w"> </span><span class="nn">sklearn.datasets</span><span class="w"> </span><span class="kn">import</span> <span class="n">make_hastie_10_2</span>
<span class="gp">>>> </span><span class="n">X</span><span class="p">,</span> <span class="n">y</span> <span class="o">=</span> <span class="n">make_hastie_10_2</span><span class="p">(</span><span class="n">random_state</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">X_train</span><span class="p">,</span> <span class="n">X_test</span> <span class="o">=</span> <span class="n">X</span><span class="p">[:</span><span class="mi">2000</span><span class="p">],</span> <span class="n">X</span><span class="p">[</span><span class="mi">2000</span><span class="p">:]</span>
<span class="gp">>>> </span><span class="n">y_train</span><span class="p">,</span> <span class="n">y_test</span> <span class="o">=</span> <span class="n">y</span><span class="p">[:</span><span class="mi">2000</span><span class="p">],</span> <span class="n">y</span><span class="p">[</span><span class="mi">2000</span><span class="p">:]</span>
<span class="gp">>>> </span><span class="n">clf</span> <span class="o">=</span> <span class="n">HistGradientBoostingClassifier</span><span class="p">(</span><span class="n">max_iter</span><span class="o">=</span><span class="mi">100</span><span class="p">)</span><span class="o">.</span><span class="n">fit</span><span class="p">(</span><span class="n">X_train</span><span class="p">,</span> <span class="n">y_train</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">clf</span><span class="o">.</span><span class="n">score</span><span class="p">(</span><span class="n">X_test</span><span class="p">,</span> <span class="n">y_test</span><span class="p">)</span>
<span class="go">0.8965</span>
</pre></div>
</div>
<p>Available losses for <strong>regression</strong> are:</p>
<ul class="simple">
<li><p>‘squared_error’, which is the default loss;</p></li>
<li><p>‘absolute_error’, which is less sensitive to outliers than the squared error;</p></li>
<li><p>‘gamma’, which is well suited to model strictly positive outcomes;</p></li>
<li><p>‘poisson’, which is well suited to model counts and frequencies;</p></li>
<li><p>‘quantile’, which allows for estimating a conditional quantile that can later
be used to obtain prediction intervals.</p></li>
</ul>
<p>For <strong>classification</strong>, ‘log_loss’ is the only option. For binary classification
it uses the binary log loss, also known as binomial deviance or binary
cross-entropy. For <code class="docutils literal notranslate"><span class="pre">n_classes</span> <span class="pre">>=</span> <span class="pre">3</span></code>, it uses the multi-class log loss function,
with multinomial deviance and categorical cross-entropy as alternative names.
The appropriate loss version is selected based on <a class="reference internal" href="../glossary.html#term-y"><span class="xref std std-term">y</span></a> passed to
<a class="reference internal" href="../glossary.html#term-fit"><span class="xref std std-term">fit</span></a>.</p>
<p>The size of the trees can be controlled through the <code class="docutils literal notranslate"><span class="pre">max_leaf_nodes</span></code>,
<code class="docutils literal notranslate"><span class="pre">max_depth</span></code>, and <code class="docutils literal notranslate"><span class="pre">min_samples_leaf</span></code> parameters.</p>
<p>The number of bins used to bin the data is controlled with the <code class="docutils literal notranslate"><span class="pre">max_bins</span></code>
parameter. Using less bins acts as a form of regularization. It is generally
recommended to use as many bins as possible (255), which is the default.</p>
<p>The <code class="docutils literal notranslate"><span class="pre">l2_regularization</span></code> parameter acts as a regularizer for the loss function,
and corresponds to <span class="math notranslate nohighlight">\(\lambda\)</span> in the following expression (see equation (2)
in <a class="reference internal" href="#xgboost" id="id4"><span>[XGBoost]</span></a>):</p>
<div class="math notranslate nohighlight">
\[\mathcal{L}(\phi) = \sum_i l(\hat{y}_i, y_i) + \frac12 \sum_k \lambda ||w_k||^2\]</div>
<details class="sd-sphinx-override sd-dropdown sd-card sd-mb-3" id="details-on-l2-regularization">
<summary class="sd-summary-title sd-card-header">
<span class="sd-summary-text">Details on l2 regularization<a class="headerlink" href="#details-on-l2-regularization" title="Link to this dropdown">#</a></span><span class="sd-summary-state-marker sd-summary-chevron-right"><svg version="1.1" width="1.5em" height="1.5em" class="sd-octicon sd-octicon-chevron-right" viewBox="0 0 24 24" aria-hidden="true"><path d="M8.72 18.78a.75.75 0 0 1 0-1.06L14.44 12 8.72 6.28a.751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018l6.25 6.25a.75.75 0 0 1 0 1.06l-6.25 6.25a.75.75 0 0 1-1.06 0Z"></path></svg></span></summary><div class="sd-summary-content sd-card-body docutils">
<p class="sd-card-text">It is important to notice that the loss term <span class="math notranslate nohighlight">\(l(\hat{y}_i, y_i)\)</span> describes
only half of the actual loss function except for the pinball loss and absolute
error.</p>
<p class="sd-card-text">The index <span class="math notranslate nohighlight">\(k\)</span> refers to the k-th tree in the ensemble of trees. In the
case of regression and binary classification, gradient boosting models grow one
tree per iteration, then <span class="math notranslate nohighlight">\(k\)</span> runs up to <code class="docutils literal notranslate"><span class="pre">max_iter</span></code>. In the case of
multiclass classification problems, the maximal value of the index <span class="math notranslate nohighlight">\(k\)</span> is
<code class="docutils literal notranslate"><span class="pre">n_classes</span></code> <span class="math notranslate nohighlight">\(\times\)</span> <code class="docutils literal notranslate"><span class="pre">max_iter</span></code>.</p>
<p class="sd-card-text">If <span class="math notranslate nohighlight">\(T_k\)</span> denotes the number of leaves in the k-th tree, then <span class="math notranslate nohighlight">\(w_k\)</span>
is a vector of length <span class="math notranslate nohighlight">\(T_k\)</span>, which contains the leaf values of the form <code class="docutils literal notranslate"><span class="pre">w</span>
<span class="pre">=</span> <span class="pre">-sum_gradient</span> <span class="pre">/</span> <span class="pre">(sum_hessian</span> <span class="pre">+</span> <span class="pre">l2_regularization)</span></code> (see equation (5) in
<a class="reference internal" href="#xgboost" id="id5"><span>[XGBoost]</span></a>).</p>
<p class="sd-card-text">The leaf values <span class="math notranslate nohighlight">\(w_k\)</span> are derived by dividing the sum of the gradients of
the loss function by the combined sum of hessians. Adding the regularization to
the denominator penalizes the leaves with small hessians (flat regions),
resulting in smaller updates. Those <span class="math notranslate nohighlight">\(w_k\)</span> values contribute then to the
model’s prediction for a given input that ends up in the corresponding leaf. The
final prediction is the sum of the base prediction and the contributions from
each tree. The result of that sum is then transformed by the inverse link
function depending on the choice of the loss function (see
<a class="reference internal" href="#gradient-boosting-formulation"><span class="std std-ref">Mathematical formulation</span></a>).</p>
<p class="sd-card-text">Notice that the original paper <a class="reference internal" href="#xgboost" id="id6"><span>[XGBoost]</span></a> introduces a term <span class="math notranslate nohighlight">\(\gamma\sum_k
T_k\)</span> that penalizes the number of leaves (making it a smooth version of
<code class="docutils literal notranslate"><span class="pre">max_leaf_nodes</span></code>) not presented here as it is not implemented in scikit-learn;
whereas <span class="math notranslate nohighlight">\(\lambda\)</span> penalizes the magnitude of the individual tree
predictions before being rescaled by the learning rate, see
<a class="reference internal" href="#gradient-boosting-shrinkage"><span class="std std-ref">Shrinkage via learning rate</span></a>.</p>
</div>
</details><p>Note that <strong>early-stopping is enabled by default if the number of samples is
larger than 10,000</strong>. The early-stopping behaviour is controlled via the
<code class="docutils literal notranslate"><span class="pre">early_stopping</span></code>, <code class="docutils literal notranslate"><span class="pre">scoring</span></code>, <code class="docutils literal notranslate"><span class="pre">validation_fraction</span></code>,
<code class="docutils literal notranslate"><span class="pre">n_iter_no_change</span></code>, and <code class="docutils literal notranslate"><span class="pre">tol</span></code> parameters. It is possible to early-stop
using an arbitrary <a class="reference internal" href="../glossary.html#term-scorer"><span class="xref std std-term">scorer</span></a>, or just the training or validation loss.
Note that for technical reasons, using a callable as a scorer is significantly slower
than using the loss. By default, early-stopping is performed if there are at least
10,000 samples in the training set, using the validation loss.</p>
</section>
<section id="missing-values-support">
<span id="nan-support-hgbt"></span><h4><span class="section-number">1.11.1.1.2. </span>Missing values support<a class="headerlink" href="#missing-values-support" title="Link to this heading">#</a></h4>
<p><a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingClassifier.html#sklearn.ensemble.HistGradientBoostingClassifier" title="sklearn.ensemble.HistGradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingClassifier</span></code></a> and
<a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingRegressor.html#sklearn.ensemble.HistGradientBoostingRegressor" title="sklearn.ensemble.HistGradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingRegressor</span></code></a> have built-in support for missing
values (NaNs).</p>
<p>During training, the tree grower learns at each split point whether samples
with missing values should go to the left or right child, based on the
potential gain. When predicting, samples with missing values are assigned to
the left or right child consequently:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="kn">from</span><span class="w"> </span><span class="nn">sklearn.ensemble</span><span class="w"> </span><span class="kn">import</span> <span class="n">HistGradientBoostingClassifier</span>
<span class="gp">>>> </span><span class="kn">import</span><span class="w"> </span><span class="nn">numpy</span><span class="w"> </span><span class="k">as</span><span class="w"> </span><span class="nn">np</span>
<span class="gp">>>> </span><span class="n">X</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">array</span><span class="p">([</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">])</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">y</span> <span class="o">=</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span>
<span class="gp">>>> </span><span class="n">gbdt</span> <span class="o">=</span> <span class="n">HistGradientBoostingClassifier</span><span class="p">(</span><span class="n">min_samples_leaf</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">fit</span><span class="p">(</span><span class="n">X</span><span class="p">,</span> <span class="n">y</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">gbdt</span><span class="o">.</span><span class="n">predict</span><span class="p">(</span><span class="n">X</span><span class="p">)</span>
<span class="go">array([0, 0, 1, 1])</span>
</pre></div>
</div>
<p>When the missingness pattern is predictive, the splits can be performed on
whether the feature value is missing or not:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">X</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">array</span><span class="p">([</span><span class="mi">0</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">])</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">y</span> <span class="o">=</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span>
<span class="gp">>>> </span><span class="n">gbdt</span> <span class="o">=</span> <span class="n">HistGradientBoostingClassifier</span><span class="p">(</span><span class="n">min_samples_leaf</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="gp">... </span> <span class="n">max_depth</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span>
<span class="gp">... </span> <span class="n">learning_rate</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="gp">... </span> <span class="n">max_iter</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">fit</span><span class="p">(</span><span class="n">X</span><span class="p">,</span> <span class="n">y</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">gbdt</span><span class="o">.</span><span class="n">predict</span><span class="p">(</span><span class="n">X</span><span class="p">)</span>
<span class="go">array([0, 1, 0, 0, 1])</span>
</pre></div>
</div>
<p>If no missing values were encountered for a given feature during training,
then samples with missing values are mapped to whichever child has the most
samples.</p>
<p class="rubric">Examples</p>
<ul class="simple">
<li><p><a class="reference internal" href="../auto_examples/ensemble/plot_hgbt_regression.html#sphx-glr-auto-examples-ensemble-plot-hgbt-regression-py"><span class="std std-ref">Features in Histogram Gradient Boosting Trees</span></a></p></li>
</ul>
</section>
<section id="sample-weight-support">
<span id="sw-hgbdt"></span><h4><span class="section-number">1.11.1.1.3. </span>Sample weight support<a class="headerlink" href="#sample-weight-support" title="Link to this heading">#</a></h4>
<p><a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingClassifier.html#sklearn.ensemble.HistGradientBoostingClassifier" title="sklearn.ensemble.HistGradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingClassifier</span></code></a> and
<a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingRegressor.html#sklearn.ensemble.HistGradientBoostingRegressor" title="sklearn.ensemble.HistGradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingRegressor</span></code></a> support sample weights during
<a class="reference internal" href="../glossary.html#term-fit"><span class="xref std std-term">fit</span></a>.</p>
<p>The following toy example demonstrates that samples with a sample weight of zero are ignored:</p>
<div class="doctest highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">X</span> <span class="o">=</span> <span class="p">[[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span>
<span class="gp">... </span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span>
<span class="gp">... </span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span>
<span class="gp">... </span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">]]</span>
<span class="gp">>>> </span><span class="n">y</span> <span class="o">=</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">]</span>
<span class="gp">>>> </span><span class="c1"># ignore the first 2 training samples by setting their weight to 0</span>
<span class="gp">>>> </span><span class="n">sample_weight</span> <span class="o">=</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span>
<span class="gp">>>> </span><span class="n">gb</span> <span class="o">=</span> <span class="n">HistGradientBoostingClassifier</span><span class="p">(</span><span class="n">min_samples_leaf</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
<span class="gp">>>> </span><span class="n">gb</span><span class="o">.</span><span class="n">fit</span><span class="p">(</span><span class="n">X</span><span class="p">,</span> <span class="n">y</span><span class="p">,</span> <span class="n">sample_weight</span><span class="o">=</span><span class="n">sample_weight</span><span class="p">)</span>
<span class="go">HistGradientBoostingClassifier(...)</span>
<span class="gp">>>> </span><span class="n">gb</span><span class="o">.</span><span class="n">predict</span><span class="p">([[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">]])</span>
<span class="go">array([1])</span>
<span class="gp">>>> </span><span class="n">gb</span><span class="o">.</span><span class="n">predict_proba</span><span class="p">([[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">]])[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span>
<span class="go">np.float64(0.999...)</span>
</pre></div>
</div>
<p>As you can see, the <code class="docutils literal notranslate"><span class="pre">[1,</span> <span class="pre">0]</span></code> is comfortably classified as <code class="docutils literal notranslate"><span class="pre">1</span></code> since the first
two samples are ignored due to their sample weights.</p>
<p>Implementation detail: taking sample weights into account amounts to
multiplying the gradients (and the hessians) by the sample weights. Note that
the binning stage (specifically the quantiles computation) does not take the
weights into account.</p>
</section>
<section id="categorical-features-support">
<span id="categorical-support-gbdt"></span><h4><span class="section-number">1.11.1.1.4. </span>Categorical Features Support<a class="headerlink" href="#categorical-features-support" title="Link to this heading">#</a></h4>
<p><a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingClassifier.html#sklearn.ensemble.HistGradientBoostingClassifier" title="sklearn.ensemble.HistGradientBoostingClassifier"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingClassifier</span></code></a> and
<a class="reference internal" href="generated/sklearn.ensemble.HistGradientBoostingRegressor.html#sklearn.ensemble.HistGradientBoostingRegressor" title="sklearn.ensemble.HistGradientBoostingRegressor"><code class="xref py py-class docutils literal notranslate"><span class="pre">HistGradientBoostingRegressor</span></code></a> have native support for categorical
features: they can consider splits on non-ordered, categorical data.</p>
<p>For datasets with categorical features, using the native categorical support
is often better than relying on one-hot encoding
(<a class="reference internal" href="generated/sklearn.preprocessing.OneHotEncoder.html#sklearn.preprocessing.OneHotEncoder" title="sklearn.preprocessing.OneHotEncoder"><code class="xref py py-class docutils literal notranslate"><span class="pre">OneHotEncoder</span></code></a>), because one-hot encoding
requires more tree depth to achieve equivalent splits. It is also usually
better to rely on the native categorical support rather than to treat
categorical features as continuous (ordinal), which happens for ordinal-encoded
categorical data, since categories are nominal quantities where order does not
matter.</p>
<p>To enable categorical support, a boolean mask can be passed to the
<code class="docutils literal notranslate"><span class="pre">categorical_features</span></code> parameter, indicating which feature is categorical. In
the following, the first feature will be treated as categorical and the
second feature as numerical:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">gbdt</span> <span class="o">=</span> <span class="n">HistGradientBoostingClassifier</span><span class="p">(</span><span class="n">categorical_features</span><span class="o">=</span><span class="p">[</span><span class="kc">True</span><span class="p">,</span> <span class="kc">False</span><span class="p">])</span>
</pre></div>
</div>
<p>Equivalently, one can pass a list of integers indicating the indices of the
categorical features:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">gbdt</span> <span class="o">=</span> <span class="n">HistGradientBoostingClassifier</span><span class="p">(</span><span class="n">categorical_features</span><span class="o">=</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
</pre></div>
</div>
<p>When the input is a DataFrame, it is also possible to pass a list of column
names:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">gbdt</span> <span class="o">=</span> <span class="n">HistGradientBoostingClassifier</span><span class="p">(</span><span class="n">categorical_features</span><span class="o">=</span><span class="p">[</span><span class="s2">"site"</span><span class="p">,</span> <span class="s2">"manufacturer"</span><span class="p">])</span>
</pre></div>
</div>
<p>Finally, when the input is a DataFrame we can use
<code class="docutils literal notranslate"><span class="pre">categorical_features="from_dtype"</span></code> in which case all columns with a categorical
<code class="docutils literal notranslate"><span class="pre">dtype</span></code> will be treated as categorical features.</p>
<p>The cardinality of each categorical feature must be less than the <code class="docutils literal notranslate"><span class="pre">max_bins</span></code>
parameter. For an example using histogram-based gradient boosting on categorical
features, see
<a class="reference internal" href="../auto_examples/ensemble/plot_gradient_boosting_categorical.html#sphx-glr-auto-examples-ensemble-plot-gradient-boosting-categorical-py"><span class="std std-ref">Categorical Feature Support in Gradient Boosting</span></a>.</p>
<p>If there are missing values during training, the missing values will be
treated as a proper category. If there are no missing values during training,
then at prediction time, missing values are mapped to the child node that has
the most samples (just like for continuous features). When predicting,
categories that were not seen during fit time will be treated as missing
values.</p>
<details class="sd-sphinx-override sd-dropdown sd-card sd-mb-3" id="split-finding-with-categorical-features">
<summary class="sd-summary-title sd-card-header">
<span class="sd-summary-text">Split finding with categorical features<a class="headerlink" href="#split-finding-with-categorical-features" title="Link to this dropdown">#</a></span><span class="sd-summary-state-marker sd-summary-chevron-right"><svg version="1.1" width="1.5em" height="1.5em" class="sd-octicon sd-octicon-chevron-right" viewBox="0 0 24 24" aria-hidden="true"><path d="M8.72 18.78a.75.75 0 0 1 0-1.06L14.44 12 8.72 6.28a.751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018l6.25 6.25a.75.75 0 0 1 0 1.06l-6.25 6.25a.75.75 0 0 1-1.06 0Z"></path></svg></span></summary><div class="sd-summary-content sd-card-body docutils">
<p class="sd-card-text">The canonical way of considering categorical splits in a tree is to consider
all of the <span class="math notranslate nohighlight">\(2^{K - 1} - 1\)</span> partitions, where <span class="math notranslate nohighlight">\(K\)</span> is the number of
categories. This can quickly become prohibitive when <span class="math notranslate nohighlight">\(K\)</span> is large.
Fortunately, since gradient boosting trees are always regression trees (even
for classification problems), there exist a faster strategy that can yield
equivalent splits. First, the categories of a feature are sorted according to
the variance of the target, for each category <code class="docutils literal notranslate"><span class="pre">k</span></code>. Once the categories are
sorted, one can consider <em>continuous partitions</em>, i.e. treat the categories
as if they were ordered continuous values (see Fisher <a class="reference internal" href="#fisher1958" id="id7"><span>[Fisher1958]</span></a> for a
formal proof). As a result, only <span class="math notranslate nohighlight">\(K - 1\)</span> splits need to be considered
instead of <span class="math notranslate nohighlight">\(2^{K - 1} - 1\)</span>. The initial sorting is a
<span class="math notranslate nohighlight">\(\mathcal{O}(K \log(K))\)</span> operation, leading to a total complexity of
<span class="math notranslate nohighlight">\(\mathcal{O}(K \log(K) + K)\)</span>, instead of <span class="math notranslate nohighlight">\(\mathcal{O}(2^K)\)</span>.</p>
</div>
</details><p class="rubric">Examples</p>
<ul class="simple">
<li><p><a class="reference internal" href="../auto_examples/ensemble/plot_gradient_boosting_categorical.html#sphx-glr-auto-examples-ensemble-plot-gradient-boosting-categorical-py"><span class="std std-ref">Categorical Feature Support in Gradient Boosting</span></a></p></li>
</ul>
</section>
<section id="monotonic-constraints">
<span id="monotonic-cst-gbdt"></span><h4><span class="section-number">1.11.1.1.5. </span>Monotonic Constraints<a class="headerlink" href="#monotonic-constraints" title="Link to this heading">#</a></h4>
<p>Depending on the problem at hand, you may have prior knowledge indicating
that a given feature should in general have a positive (or negative) effect
on the target value. For example, all else being equal, a higher credit
score should increase the probability of getting approved for a loan.
Monotonic constraints allow you to incorporate such prior knowledge into the
model.</p>
<p>For a predictor <span class="math notranslate nohighlight">\(F\)</span> with two features:</p>
<ul>
<li><p>a <strong>monotonic increase constraint</strong> is a constraint of the form:</p>
<div class="math notranslate nohighlight">
\[x_1 \leq x_1' \implies F(x_1, x_2) \leq F(x_1', x_2)\]</div>
</li>
<li><p>a <strong>monotonic decrease constraint</strong> is a constraint of the form:</p>
<div class="math notranslate nohighlight">
\[x_1 \leq x_1' \implies F(x_1, x_2) \geq F(x_1', x_2)\]</div>
</li>
</ul>
<p>You can specify a monotonic constraint on each feature using the
<code class="docutils literal notranslate"><span class="pre">monotonic_cst</span></code> parameter. For each feature, a value of 0 indicates no
constraint, while 1 and -1 indicate a monotonic increase and
monotonic decrease constraint, respectively:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="kn">from</span><span class="w"> </span><span class="nn">sklearn.ensemble</span><span class="w"> </span><span class="kn">import</span> <span class="n">HistGradientBoostingRegressor</span>
<span class="go">... # monotonic increase, monotonic decrease, and no constraint on the 3 features</span>
<span class="gp">>>> </span><span class="n">gbdt</span> <span class="o">=</span> <span class="n">HistGradientBoostingRegressor</span><span class="p">(</span><span class="n">monotonic_cst</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">])</span>
</pre></div>
</div>
<p>In a binary classification context, imposing a monotonic increase (decrease) constraint means that higher values of the feature are supposed
to have a positive (negative) effect on the probability of samples
to belong to the positive class.</p>
<p>Nevertheless, monotonic constraints only marginally constrain feature effects on the output.