|
1
|
|
|
<?php |
|
2
|
|
|
declare(strict_types=1); |
|
3
|
|
|
|
|
4
|
|
|
namespace Phpml\Classification\Ensemble; |
|
5
|
|
|
|
|
6
|
|
|
use Phpml\Classification\Ensemble\Bagging; |
|
7
|
|
|
use Phpml\Classification\DecisionTree; |
|
8
|
|
|
use Phpml\Classification\NaiveBayes; |
|
9
|
|
|
use Phpml\Classification\Classifier; |
|
10
|
|
|
|
|
11
|
|
|
class RandomForest extends Bagging |
|
12
|
|
|
{ |
|
13
|
|
|
/** |
|
14
|
|
|
* @var float |
|
15
|
|
|
*/ |
|
16
|
|
|
protected $subsetRatio = 1.0; |
|
17
|
|
|
|
|
18
|
|
|
/** |
|
19
|
|
|
* @var float|string |
|
20
|
|
|
*/ |
|
21
|
|
|
protected $featureSubsetRatio = 'log'; |
|
22
|
|
|
|
|
23
|
|
|
/** |
|
24
|
|
|
* This method is used to determine how much of the original columns (features) |
|
25
|
|
|
* will be used to construct subsets to train base classifiers.<br> |
|
26
|
|
|
* |
|
27
|
|
|
* Allowed values: 'sqrt', 'log' or any float number between 0.1 and 1.0 <br> |
|
28
|
|
|
* |
|
29
|
|
|
* If there are many features that diminishes classification performance, then |
|
30
|
|
|
* small values should be preferred, otherwise, with low number of features, |
|
31
|
|
|
* default value (0.7) will result in satisfactory performance. |
|
32
|
|
|
* |
|
33
|
|
|
* @param mixed $ratio string or float should be given |
|
34
|
|
|
* @return $this |
|
35
|
|
|
* @throws Exception |
|
36
|
|
|
*/ |
|
37
|
|
|
public function setFeatureSubsetRatio($ratio) |
|
38
|
|
|
{ |
|
39
|
|
|
if (is_float($ratio) && ($ratio < 0.1 || $ratio > 1.0)) { |
|
40
|
|
|
throw new \Exception("When a float given, feature subset ratio should be between 0.1 and 1.0"); |
|
41
|
|
|
} |
|
42
|
|
|
if (is_string($ratio) && $ratio != 'sqrt' && $ratio != 'log') { |
|
43
|
|
|
throw new \Exception("When a string given, feature subset ratio can only be 'sqrt' or 'log' "); |
|
44
|
|
|
} |
|
45
|
|
|
$this->featureSubsetRatio = $ratio; |
|
46
|
|
|
return $this; |
|
47
|
|
|
} |
|
48
|
|
|
|
|
49
|
|
|
/** |
|
50
|
|
|
* RandomForest algorithm is usable *only* with DecisionTree |
|
51
|
|
|
* |
|
52
|
|
|
* @param string $classifier |
|
53
|
|
|
* @param array $classifierOptions |
|
54
|
|
|
* @return $this |
|
55
|
|
|
*/ |
|
56
|
|
|
public function setClassifer(string $classifier, array $classifierOptions = []) |
|
57
|
|
|
{ |
|
58
|
|
|
if ($classifier != DecisionTree::class) { |
|
59
|
|
|
throw new \Exception("RandomForest can only use DecisionTree as base classifier"); |
|
60
|
|
|
} |
|
61
|
|
|
|
|
62
|
|
|
return parent::setClassifer($classifier, $classifierOptions); |
|
63
|
|
|
} |
|
64
|
|
|
|
|
65
|
|
|
/** |
|
66
|
|
|
* @param DecisionTree $classifier |
|
67
|
|
|
* @param int $index |
|
68
|
|
|
* @return DecisionTree |
|
69
|
|
|
*/ |
|
70
|
|
|
protected function initSingleClassifier($classifier, $index) |
|
71
|
|
|
{ |
|
72
|
|
|
if (is_float($this->featureSubsetRatio)) { |
|
73
|
|
|
$featureCount = (int)($this->featureSubsetRatio * $this->featureCount); |
|
74
|
|
|
} elseif ($this->featureCount == 'sqrt') { |
|
75
|
|
|
$featureCount = (int)sqrt($this->featureCount) + 1; |
|
76
|
|
|
} else { |
|
77
|
|
|
$featureCount = (int)log($this->featureCount, 2) + 1; |
|
78
|
|
|
} |
|
79
|
|
|
|
|
80
|
|
|
if ($featureCount >= $this->featureCount) { |
|
81
|
|
|
$featureCount = $this->featureCount; |
|
82
|
|
|
} |
|
83
|
|
|
|
|
84
|
|
|
return $classifier->setNumFeatures($featureCount); |
|
85
|
|
|
} |
|
86
|
|
|
} |
|
87
|
|
|
|
php-ai /
php-ml
