-
-
Notifications
You must be signed in to change notification settings - Fork 195
Expand file tree
/
Copy pathRandomForest.php
More file actions
371 lines (313 loc) · 9.7 KB
/
RandomForest.php
File metadata and controls
371 lines (313 loc) · 9.7 KB
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
<?php
namespace Rubix\ML\Classifiers;
use Rubix\ML\Learner;
use Rubix\ML\Parallel;
use Rubix\ML\Estimator;
use Rubix\ML\Persistable;
use Rubix\ML\Probabilistic;
use Rubix\ML\RanksFeatures;
use Rubix\ML\EstimatorType;
use Rubix\ML\Helpers\Params;
use Rubix\ML\Backends\Serial;
use Rubix\ML\Datasets\Dataset;
use Rubix\ML\Backends\Tasks\Proba;
use Rubix\ML\Traits\Multiprocessing;
use Rubix\ML\Backends\Tasks\Predict;
use Rubix\ML\Backends\Tasks\TrainLearner;
use Rubix\ML\Traits\AutotrackRevisions;
use Rubix\ML\Specifications\DatasetIsLabeled;
use Rubix\ML\Specifications\DatasetIsNotEmpty;
use Rubix\ML\Specifications\SpecificationChain;
use Rubix\ML\Specifications\DatasetHasDimensionality;
use Rubix\ML\Specifications\LabelsAreCompatibleWithLearner;
use Rubix\ML\Specifications\SamplesAreCompatibleWithEstimator;
use Rubix\ML\Exceptions\InvalidArgumentException;
use Rubix\ML\Exceptions\RuntimeException;
use function Rubix\ML\argmax;
use function Rubix\ML\array_transpose;
use function array_count_values;
use function get_class;
use function in_array;
/**
* Random Forest
*
* An ensemble classifier that trains an ensemble of Decision Trees (Classification or Extra Trees)
* on random subsets (*bootstrap* set) of the training data. Predictions are based on the
* probability scores returned from each tree in the forest, averaged and weighted equally.
*
* References:
* [1] L. Breiman. (2001). Random Forests.
* [2] L. Breiman et al. (2005). Extremely Randomized Trees.
*
* @category Machine Learning
* @package Rubix/ML
* @author Andrew DalPino
*/
class RandomForest implements Estimator, Learner, Probabilistic, Parallel, RanksFeatures, Persistable
{
use AutotrackRevisions, Multiprocessing;
/**
* The class names of the learners that are compatible with the ensemble.
*
* @var class-string[]
*/
public const COMPATIBLE_LEARNERS = [
ClassificationTree::class,
ExtraTreeClassifier::class,
];
/**
* The minimum size of each training subset.
*
* @var int
*/
protected const MIN_SUBSAMPLE = 1;
/**
* The base learner.
*
* @var Learner
*/
protected Learner $base;
/**
* The number of learners to train in the ensemble.
*
* @var int
*/
protected int $estimators;
/**
* The ratio of samples from the training set to randomly subsample to train each base learner.
*
* @var float
*/
protected float $ratio;
/**
* Should we sample the bootstrap set to compensate for imbalanced class labels?
*
* @var bool
*/
protected bool $balanced;
/**
* The decision trees that make up the forest.
*
* @var list<ClassificationTree|ExtraTreeClassifier>|null
*/
protected ?array $trees = null;
/**
* The zero vector for the possible class outcomes.
*
* @var float[]|null
*/
protected ?array $classes = null;
/**
* The dimensionality of the training set.
*
* @var int<0,max>|null
*/
protected ?int $featureCount = null;
/**
* @param Learner|null $base
* @param int $estimators
* @param float $ratio
* @param bool $balanced
* @throws InvalidArgumentException
*/
public function __construct(
?Learner $base = null,
int $estimators = 100,
float $ratio = 0.2,
bool $balanced = false
) {
if ($base and !in_array(get_class($base), self::COMPATIBLE_LEARNERS)) {
throw new InvalidArgumentException('Base Learner must be'
. ' compatible with ensemble.');
}
if ($estimators < 1) {
throw new InvalidArgumentException('Number of estimators'
. " must be greater than 0, $estimators given.");
}
if ($ratio <= 0.0 or $ratio > 1.5) {
throw new InvalidArgumentException('Ratio must be between'
. " 0 and 1.5, $ratio given.");
}
$this->base = $base ?? new ClassificationTree();
$this->estimators = $estimators;
$this->ratio = $ratio;
$this->balanced = $balanced;
$this->backend = new Serial();
}
/**
* Return the estimator type.
*
* @internal
*
* @return EstimatorType
*/
public function type() : EstimatorType
{
return EstimatorType::classifier();
}
/**
* Return the data types that the estimator is compatible with.
*
* @internal
*
* @return list<\Rubix\ML\DataType>
*/
public function compatibility() : array
{
return $this->base->compatibility();
}
/**
* Return the settings of the hyper-parameters in an associative array.
*
* @internal
*
* @return mixed[]
*/
public function params() : array
{
return [
'base' => $this->base,
'estimators' => $this->estimators,
'ratio' => $this->ratio,
'balanced' => $this->balanced,
];
}
/**
* Has the learner been trained?
*
* @return bool
*/
public function trained() : bool
{
return !empty($this->trees);
}
/**
* Train the learner with a dataset.
*
* @param \Rubix\ML\Datasets\Labeled $dataset
*/
public function train(Dataset $dataset) : void
{
SpecificationChain::with([
new DatasetIsLabeled($dataset),
new DatasetIsNotEmpty($dataset),
new SamplesAreCompatibleWithEstimator($dataset, $this),
new LabelsAreCompatibleWithLearner($dataset, $this),
])->check();
$p = max(self::MIN_SUBSAMPLE, (int) ceil($this->ratio * $dataset->numSamples()));
if ($this->balanced) {
$counts = array_count_values($dataset->labels());
$min = min($counts);
$weights = [];
foreach ($dataset->labels() as $label) {
$weights[] = $min / $counts[$label];
}
}
$this->backend->flush();
for ($i = 0; $i < $this->estimators; ++$i) {
$estimator = clone $this->base;
if (isset($weights)) {
$subset = $dataset->randomWeightedSubsetWithReplacement($p, $weights);
} else {
$subset = $dataset->randomSubsetWithReplacement($p);
}
$this->backend->enqueue(new TrainLearner($estimator, $subset));
}
$this->trees = $this->backend->process();
$this->classes = array_fill_keys($dataset->possibleOutcomes(), 0.0);
$this->featureCount = $dataset->numFeatures();
}
/**
* Make predictions from a dataset.
*
* @param Dataset $dataset
* @throws RuntimeException
* @return list<string>
*/
public function predict(Dataset $dataset) : array
{
if (!$this->trees or !$this->featureCount) {
throw new RuntimeException('Estimator has not been trained.');
}
DatasetHasDimensionality::with($dataset, $this->featureCount)->check();
$this->backend->flush();
foreach ($this->trees as $estimator) {
$this->backend->enqueue(new Predict($estimator, $dataset));
}
$aggregate = array_transpose($this->backend->process());
$predictions = [];
foreach ($aggregate as $votes) {
/** @var array<string,int> $counts */
$counts = array_count_values($votes);
$predictions[] = argmax($counts);
}
return $predictions;
}
/**
* Estimate the joint probabilities for each possible outcome.
*
* @param Dataset $dataset
* @throws RuntimeException
* @return list<array<string,float>>
*/
public function proba(Dataset $dataset) : array
{
if (!$this->trees or !$this->classes or !$this->featureCount) {
throw new RuntimeException('Estimator has not been trained.');
}
DatasetHasDimensionality::with($dataset, $this->featureCount)->check();
$probabilities = array_fill(0, $dataset->numSamples(), $this->classes);
$this->backend->flush();
foreach ($this->trees as $estimator) {
$this->backend->enqueue(new Proba($estimator, $dataset));
}
$aggregate = $this->backend->process();
foreach ($aggregate as $proba) {
/** @var int $i */
foreach ($proba as $i => $joint) {
foreach ($joint as $class => $probability) {
$probabilities[$i][$class] += $probability;
}
}
}
foreach ($probabilities as &$joint) {
foreach ($joint as &$probability) {
$probability /= $this->estimators;
}
}
return $probabilities;
}
/**
* Return the importance scores of each feature column of the training set.
*
* @throws RuntimeException
* @return float[]
*/
public function featureImportances() : array
{
if (!$this->trees or !$this->featureCount) {
throw new RuntimeException('Estimator has not been trained.');
}
$importances = array_fill(0, $this->featureCount, 0.0);
foreach ($this->trees as $tree) {
foreach ($tree->featureImportances() as $column => $importance) {
$importances[$column] += $importance;
}
}
foreach ($importances as &$importance) {
$importance /= $this->estimators;
}
return $importances;
}
/**
* Return the string representation of the object.
*
* @internal
*
* @return string
*/
public function __toString() : string
{
return 'Random Forest (' . Params::stringify($this->params()) . ')';
}
}