trepan-demo.ipynb 31.1 KB
Newer Older
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
{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using Trepan: a demonstration\n",
    "\n",
    "By: Yuriy Sverchkov\n",
    "\n",
    "Assistant Scientist\n",
    "\n",
    "Department of Biostatistics and Medical Informatics\n",
    "\n",
    "University of Wisconsin - Madison, USA\n",
    "\n",
    "## Trepan\n",
    "\n",
    "[M. Craven and J. Shalvik. 1995. Extracting tree-structured representations of trained networks. In Proceedings of the 8th International Conference on Neural Information Processing 1995](https://dl.acm.org/doi/10.5555/2998828.2998832)\n",
    "\n",
    "Trepan is one of the first explanation-by-model-translation methods for Neural Networks.\n",
    "Model translation is an approach to model explanation in which an uninterpretable black-box model is translated into an interpretable model (in this case a decision tree) that aims at having high fidelity to the black-box model, that is, make similar prediction to it.\n",
    "\n",
    "In this notebook we will use a modern implementation in the [`generalizedtrees`](https://github.com/Craven-Biostat-Lab/generalizedtrees) package to run Trepan on one of the original datasets used in the paper. `generalizedtrees` is under development with major changes from version to version. It is a project that attempts to bring together many different variants of tree learning together into a single framework.\n",
    "\n",
    "For this notebook we will use `generalizedtrees` version 1.1.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: generalizedtrees==1.1.0 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (1.1.0)\n",
      "Requirement already satisfied: scipy>=1.5.2 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from generalizedtrees==1.1.0) (1.6.0)\n",
      "Requirement already satisfied: pandas>=1.1.0 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from generalizedtrees==1.1.0) (1.2.2)\n",
      "Requirement already satisfied: scikit-learn>=0.23.2 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from generalizedtrees==1.1.0) (0.24.1)\n",
      "Requirement already satisfied: numpy>=1.19.1 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from generalizedtrees==1.1.0) (1.20.1)\n",
      "Requirement already satisfied: python-dateutil>=2.7.3 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from pandas>=1.1.0->generalizedtrees==1.1.0) (2.8.1)\n",
      "Requirement already satisfied: pytz>=2017.3 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from pandas>=1.1.0->generalizedtrees==1.1.0) (2021.1)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from scikit-learn>=0.23.2->generalizedtrees==1.1.0) (2.1.0)\n",
      "Requirement already satisfied: joblib>=0.11 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from scikit-learn>=0.23.2->generalizedtrees==1.1.0) (1.0.1)\n",
      "Requirement already satisfied: six>=1.5 in /Users/sees/.virtualenvs/trepan-demo/lib/python3.8/site-packages (from python-dateutil>=2.7.3->pandas>=1.1.0->generalizedtrees==1.1.0) (1.15.0)\n",
      "\u001b[33mWARNING: You are using pip version 20.2.3; however, version 21.0.1 is available.\n",
      "You should consider upgrading via the '/Users/sees/.virtualenvs/trepan-demo/bin/python -m pip install --upgrade pip' command.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "!{sys.executable} -m pip install generalizedtrees==1.1.0"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data\n",
    "\n",
    "We will use the Cleveland Heart Disease dataset, available from the UCI repository: http://archive.ics.uci.edu/ml/datasets/Heart+Disease.\n",
    "The file we load here is specifically http://archive.ics.uci.edu/ml/machine-learning-databases/heart-disease/cleve.mod."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.preprocessing import OneHotEncoder, LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "np_rng = np.random.default_rng(8372234)\n",
    "sk_rng = np.random.RandomState(3957458)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>sex</th>\n",
       "      <th>chest pain type</th>\n",
       "      <th>resting bp</th>\n",
       "      <th>cholesterol</th>\n",
       "      <th>fasting blood sugar &lt; 120</th>\n",
       "      <th>resting ecg</th>\n",
       "      <th>max heart rate</th>\n",
       "      <th>exercise induced angina</th>\n",
       "      <th>oldpeak</th>\n",
       "      <th>slope</th>\n",
       "      <th>number of vessels colored</th>\n",
       "      <th>thal</th>\n",
       "      <th>class</th>\n",
       "      <th>stage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>63.0</td>\n",
       "      <td>male</td>\n",
       "      <td>angina</td>\n",
       "      <td>145.0</td>\n",
       "      <td>233.0</td>\n",
       "      <td>True</td>\n",
       "      <td>hyp</td>\n",
       "      <td>150.0</td>\n",
       "      <td>False</td>\n",
       "      <td>2.3</td>\n",
       "      <td>down</td>\n",
       "      <td>0.0</td>\n",
       "      <td>fix</td>\n",
       "      <td>buff</td>\n",
       "      <td>H</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>67.0</td>\n",
       "      <td>male</td>\n",
       "      <td>asympt</td>\n",
       "      <td>160.0</td>\n",
       "      <td>286.0</td>\n",
       "      <td>False</td>\n",
       "      <td>hyp</td>\n",
       "      <td>108.0</td>\n",
       "      <td>True</td>\n",
       "      <td>1.5</td>\n",
       "      <td>flat</td>\n",
       "      <td>3.0</td>\n",
       "      <td>norm</td>\n",
       "      <td>sick</td>\n",
       "      <td>S2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>67.0</td>\n",
       "      <td>male</td>\n",
       "      <td>asympt</td>\n",
       "      <td>120.0</td>\n",
       "      <td>229.0</td>\n",
       "      <td>False</td>\n",
       "      <td>hyp</td>\n",
       "      <td>129.0</td>\n",
       "      <td>True</td>\n",
       "      <td>2.6</td>\n",
       "      <td>flat</td>\n",
       "      <td>2.0</td>\n",
       "      <td>rev</td>\n",
       "      <td>sick</td>\n",
       "      <td>S1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>37.0</td>\n",
       "      <td>male</td>\n",
       "      <td>notang</td>\n",
       "      <td>130.0</td>\n",
       "      <td>250.0</td>\n",
       "      <td>False</td>\n",
       "      <td>norm</td>\n",
       "      <td>187.0</td>\n",
       "      <td>False</td>\n",
       "      <td>3.5</td>\n",
       "      <td>down</td>\n",
       "      <td>0.0</td>\n",
       "      <td>norm</td>\n",
       "      <td>buff</td>\n",
       "      <td>H</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>41.0</td>\n",
       "      <td>fem</td>\n",
       "      <td>abnang</td>\n",
       "      <td>130.0</td>\n",
       "      <td>204.0</td>\n",
       "      <td>False</td>\n",
       "      <td>hyp</td>\n",
       "      <td>172.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1.4</td>\n",
       "      <td>up</td>\n",
       "      <td>0.0</td>\n",
       "      <td>norm</td>\n",
       "      <td>buff</td>\n",
       "      <td>H</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>298</th>\n",
       "      <td>48.0</td>\n",
       "      <td>male</td>\n",
       "      <td>notang</td>\n",
       "      <td>124.0</td>\n",
       "      <td>255.0</td>\n",
       "      <td>True</td>\n",
       "      <td>norm</td>\n",
       "      <td>175.0</td>\n",
       "      <td>False</td>\n",
       "      <td>0.0</td>\n",
       "      <td>up</td>\n",
       "      <td>2.0</td>\n",
       "      <td>norm</td>\n",
       "      <td>buff</td>\n",
       "      <td>H</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>299</th>\n",
       "      <td>57.0</td>\n",
       "      <td>male</td>\n",
       "      <td>asympt</td>\n",
       "      <td>132.0</td>\n",
       "      <td>207.0</td>\n",
       "      <td>False</td>\n",
       "      <td>norm</td>\n",
       "      <td>168.0</td>\n",
       "      <td>True</td>\n",
       "      <td>0.0</td>\n",
       "      <td>up</td>\n",
       "      <td>0.0</td>\n",
       "      <td>rev</td>\n",
       "      <td>buff</td>\n",
       "      <td>H</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>300</th>\n",
       "      <td>49.0</td>\n",
       "      <td>male</td>\n",
       "      <td>notang</td>\n",
       "      <td>118.0</td>\n",
       "      <td>149.0</td>\n",
       "      <td>False</td>\n",
       "      <td>hyp</td>\n",
       "      <td>126.0</td>\n",
       "      <td>False</td>\n",
       "      <td>0.8</td>\n",
       "      <td>up</td>\n",
       "      <td>3.0</td>\n",
       "      <td>norm</td>\n",
       "      <td>sick</td>\n",
       "      <td>S1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>301</th>\n",
       "      <td>74.0</td>\n",
       "      <td>fem</td>\n",
       "      <td>abnang</td>\n",
       "      <td>120.0</td>\n",
       "      <td>269.0</td>\n",
       "      <td>False</td>\n",
       "      <td>hyp</td>\n",
       "      <td>121.0</td>\n",
       "      <td>True</td>\n",
       "      <td>0.2</td>\n",
       "      <td>up</td>\n",
       "      <td>1.0</td>\n",
       "      <td>norm</td>\n",
       "      <td>buff</td>\n",
       "      <td>H</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>302</th>\n",
       "      <td>54.0</td>\n",
       "      <td>fem</td>\n",
       "      <td>notang</td>\n",
       "      <td>160.0</td>\n",
       "      <td>201.0</td>\n",
       "      <td>False</td>\n",
       "      <td>norm</td>\n",
       "      <td>163.0</td>\n",
       "      <td>False</td>\n",
       "      <td>0.0</td>\n",
       "      <td>up</td>\n",
       "      <td>1.0</td>\n",
       "      <td>norm</td>\n",
       "      <td>buff</td>\n",
       "      <td>H</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>296 rows × 15 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      age   sex chest pain type  resting bp  cholesterol  \\\n",
       "0    63.0  male          angina       145.0        233.0   \n",
       "1    67.0  male          asympt       160.0        286.0   \n",
       "2    67.0  male          asympt       120.0        229.0   \n",
       "3    37.0  male          notang       130.0        250.0   \n",
       "4    41.0   fem          abnang       130.0        204.0   \n",
       "..    ...   ...             ...         ...          ...   \n",
       "298  48.0  male          notang       124.0        255.0   \n",
       "299  57.0  male          asympt       132.0        207.0   \n",
       "300  49.0  male          notang       118.0        149.0   \n",
       "301  74.0   fem          abnang       120.0        269.0   \n",
       "302  54.0   fem          notang       160.0        201.0   \n",
       "\n",
       "     fasting blood sugar < 120 resting ecg  max heart rate  \\\n",
       "0                         True         hyp           150.0   \n",
       "1                        False         hyp           108.0   \n",
       "2                        False         hyp           129.0   \n",
       "3                        False        norm           187.0   \n",
       "4                        False         hyp           172.0   \n",
       "..                         ...         ...             ...   \n",
       "298                       True        norm           175.0   \n",
       "299                      False        norm           168.0   \n",
       "300                      False         hyp           126.0   \n",
       "301                      False         hyp           121.0   \n",
       "302                      False        norm           163.0   \n",
       "\n",
       "     exercise induced angina  oldpeak slope  number of vessels colored  thal  \\\n",
       "0                      False      2.3  down                        0.0   fix   \n",
       "1                       True      1.5  flat                        3.0  norm   \n",
       "2                       True      2.6  flat                        2.0   rev   \n",
       "3                      False      3.5  down                        0.0  norm   \n",
       "4                      False      1.4    up                        0.0  norm   \n",
       "..                       ...      ...   ...                        ...   ...   \n",
       "298                    False      0.0    up                        2.0  norm   \n",
       "299                     True      0.0    up                        0.0   rev   \n",
       "300                    False      0.8    up                        3.0  norm   \n",
       "301                     True      0.2    up                        1.0  norm   \n",
       "302                    False      0.0    up                        1.0  norm   \n",
       "\n",
       "    class stage  \n",
       "0    buff     H  \n",
       "1    sick    S2  \n",
       "2    sick    S1  \n",
       "3    buff     H  \n",
       "4    buff     H  \n",
       "..    ...   ...  \n",
       "298  buff     H  \n",
       "299  buff     H  \n",
       "300  sick    S1  \n",
       "301  buff     H  \n",
       "302  buff     H  \n",
       "\n",
       "[296 rows x 15 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_data = pd.read_fwf(\n",
    "    'cleveland.txt',\n",
    "    skiprows=20,\n",
    "    names = [\n",
    "        'age', 'sex', 'chest pain type', 'resting bp', 'cholesterol', 'fasting blood sugar < 120', 'resting ecg',\n",
    "        'max heart rate', 'exercise induced angina', 'oldpeak', 'slope', 'number of vessels colored', 'thal', 'class', 'stage'\n",
    "    ],\n",
    "    true_values = ['true'],\n",
    "    false_values = ['fal'],\n",
    "    na_values = '?').dropna(axis=0)\n",
    "\n",
    "full_data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since we will be using scikit-learn to learn our black-box model, we need to convert categorical variables to numeric vectors.\n",
    "We will also make a train-test split."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_df = full_data.drop(['class', 'stage'], axis=1)\n",
    "\n",
    "encoder = OneHotEncoder(drop = 'if_binary')\n",
    "lencoder = LabelEncoder()\n",
    "\n",
    "numeric_features = data_df.select_dtypes(include = 'number')\n",
    "categorical_features_df = data_df.select_dtypes(exclude = 'number')\n",
    "categorical_features = encoder.fit_transform(categorical_features_df).toarray()\n",
    "feature_names = np.append(numeric_features.columns, encoder.get_feature_names(categorical_features_df.columns))\n",
    "\n",
    "x = np.append(\n",
    "    numeric_features,\n",
    "    categorical_features,\n",
    "    axis = 1)\n",
    "\n",
    "y = lencoder.fit_transform(full_data['class'])\n",
    "\n",
    "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.1, random_state = sk_rng)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our new feature set is below. Subsequently features will be referenced by their 0-index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Feature Name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>age</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>resting bp</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>cholesterol</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>max heart rate</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>oldpeak</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>number of vessels colored</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>sex_male</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>chest pain type_abnang</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>chest pain type_angina</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>chest pain type_asympt</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>chest pain type_notang</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>fasting blood sugar &lt; 120_True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>resting ecg_abn</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>resting ecg_hyp</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>resting ecg_norm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>exercise induced angina_True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>slope_down</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>slope_flat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>slope_up</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>thal_fix</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>thal_norm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>thal_rev</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                      Feature Name\n",
       "0                              age\n",
       "1                       resting bp\n",
       "2                      cholesterol\n",
       "3                   max heart rate\n",
       "4                          oldpeak\n",
       "5        number of vessels colored\n",
       "6                         sex_male\n",
       "7           chest pain type_abnang\n",
       "8           chest pain type_angina\n",
       "9           chest pain type_asympt\n",
       "10          chest pain type_notang\n",
       "11  fasting blood sugar < 120_True\n",
       "12                 resting ecg_abn\n",
       "13                 resting ecg_hyp\n",
       "14                resting ecg_norm\n",
       "15    exercise induced angina_True\n",
       "16                      slope_down\n",
       "17                      slope_flat\n",
       "18                        slope_up\n",
       "19                        thal_fix\n",
       "20                       thal_norm\n",
       "21                        thal_rev"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame({'Feature Name': feature_names})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Black-box model\n",
    "\n",
    "Like in the paper, our black-box model will be a fully connected Neural Network with one hidden layer, and the size of the layer is determined by cross-validation within the training set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Pipeline(steps=[('standardscaler', StandardScaler()),\n",
       "                ('mlpclassifier',\n",
       "                 MLPClassifier(alpha=1e-05, hidden_layer_sizes=(5,),\n",
       "                               random_state=RandomState(MT19937) at 0x13B59DA40,\n",
       "                               solver='lbfgs'))])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "model = GridSearchCV(\n",
    "    make_pipeline(StandardScaler(), MLPClassifier(solver='lbfgs', alpha=1e-5, random_state=sk_rng)),\n",
    "    param_grid = {'mlpclassifier__hidden_layer_sizes': [(5,), (10,), (20,), (40,)]},\n",
    "    refit = True\n",
    ")\n",
    "\n",
    "model.fit(x_train, y_train)\n",
    "\n",
    "model.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Decision tree explanation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "\n",
    "from generalizedtrees.recipes import trepan\n",
    "from generalizedtrees.vis.vis import explanation_to_html\n",
    "from generalizedtrees.features import FeatureSpec"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have a function that serves as a 'recipe' for Trepan in our `generalizedtrees` python package, this returns an object that can be fit to data and a model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "explanation = trepan(\n",
    "    m_of_n=False,\n",
    "    max_tree_size=10,\n",
    "    impurity='entropy',\n",
    "    rng = np_rng)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We learn the explanation from the black-box model using the `fit` method, which takes unlabeled data and the black-box model as input."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Assuming continuous features in the absence of feature specifications\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time taken to learn explanation: 3.302284002304077 seconds\n"
     ]
    }
   ],
   "source": [
    "t0 = time.time()\n",
    "\n",
    "explanation.fit(x_train, model)\n",
    "\n",
    "t1 = time.time()\n",
    "\n",
    "print(f'Time taken to learn explanation: {t1-t0} seconds')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A console-friendly representation of the tree is available:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test x[5] > 0.5\n",
      "+--Test x[5] > 1.5\n",
      "|  +--[0.49 0.51]\n",
      "|  +--[0.577 0.423]\n",
      "+--Test x[4] > 1.55\n",
      "   +--[0.539 0.461]\n",
      "   +--Test x[20] > 0.5\n",
      "      +--Test x[4] > 1.45\n",
      "      |  +--[0.561 0.439]\n",
      "      |  +--[0.699 0.301]\n",
      "      +--[0.601 0.399]\n"
     ]
    }
   ],
   "source": [
    "print(explanation.show_tree())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A graphical representation (as an HTML file) can be generated as well:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "explanation_to_html(explanation, 'explanation.html')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Open the generated html:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Click the follwing link: [explanation.html](explanation.html)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Performance comparison\n",
    "\n",
    "The resulting tree can itself be treated as a classifier, and we can check how this learned tree performs on test data, alongside the original black-box model for comparison."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test_trepan = explanation.predict(x_test)\n",
    "y_test_model = model.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Trepan:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        buff       0.71      0.94      0.81        16\n",
      "        sick       0.89      0.57      0.70        14\n",
      "\n",
      "    accuracy                           0.77        30\n",
      "   macro avg       0.80      0.75      0.75        30\n",
      "weighted avg       0.80      0.77      0.76        30\n",
      "\n",
      "Black Box:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        buff       0.79      0.94      0.86        16\n",
      "        sick       0.91      0.71      0.80        14\n",
      "\n",
      "    accuracy                           0.83        30\n",
      "   macro avg       0.85      0.83      0.83        30\n",
      "weighted avg       0.85      0.83      0.83        30\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "\n",
    "print('Trepan:')\n",
    "print(classification_report(y_test, y_test_trepan, target_names=lencoder.classes_))\n",
    "\n",
    "print('Black Box:')\n",
    "print(classification_report(y_test, y_test_model, target_names=lencoder.classes_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "Another metric of interest for explanation methods specifically is *fidelity*, which measures how well the explanation's predictions match the black-box predictions:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set fidelity\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        buff       0.65      0.93      0.77       151\n",
      "        sick       0.80      0.34      0.48       115\n",
      "\n",
      "    accuracy                           0.68       266\n",
      "   macro avg       0.72      0.64      0.62       266\n",
      "weighted avg       0.71      0.68      0.64       266\n",
      "\n",
      "Test set fidelity\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        buff       0.81      0.89      0.85        19\n",
      "        sick       0.78      0.64      0.70        11\n",
      "\n",
      "    accuracy                           0.80        30\n",
      "   macro avg       0.79      0.77      0.78        30\n",
      "weighted avg       0.80      0.80      0.80        30\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('Training set fidelity')\n",
    "print(classification_report(model.predict(x_train), explanation.predict(x_train), target_names=lencoder.classes_))\n",
    "\n",
    "print('Test set fidelity')\n",
    "print(classification_report(y_test_model, y_test_trepan, target_names=lencoder.classes_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Remarks\n",
    "\n",
    "There are some differences between the experiment presented here and the experiment in the paper. The version of Trepan we ran here did not use m-of-n splits at nodes (the search for splits becomes very slow if we enable it), we used a different number for the minimum samples required at each node (1000), and we used a simpler rejection-based sampling scheme for generating synthetic samples. The details of the neural network learning algorithm are also different.\n",
    "\n",
    "This notebook is available [in a gist](https://gist.github.com/sverchkov/c87b301db1b88e0f4cc8bb7d77b889b9).\n",
    "\n",
    "For issues/questions about the `generalizedtrees` package contact us through https://github.com/Craven-Biostat-Lab/generalizedtrees"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}