|
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
|
|
|
|
|
10
|
|
|
class RandomForest extends Bagging |
|
11
|
|
|
{ |
|
12
|
|
|
/** |
|
13
|
|
|
* @var array |
|
14
|
|
|
*/ |
|
15
|
|
|
protected $classifierColumns; |
|
16
|
|
|
|
|
17
|
|
|
/** |
|
18
|
|
|
* @var float |
|
19
|
|
|
*/ |
|
20
|
|
|
protected $subsetRatio = 1.0; |
|
21
|
|
|
|
|
22
|
|
|
/** |
|
23
|
|
|
* @var float|string |
|
24
|
|
|
*/ |
|
25
|
|
|
protected $featureSubsetRatio = 0.7; |
|
26
|
|
|
|
|
27
|
|
|
/** |
|
28
|
|
|
* This method is used to determine how much of the original columns (features) |
|
29
|
|
|
* will be used to construct subsets to train base classifiers.<br> |
|
30
|
|
|
* |
|
31
|
|
|
* Allowed values: 'sqrt', 'log' or any float number between 0.1 and 1.0 <br> |
|
32
|
|
|
* |
|
33
|
|
|
* If there are many features that diminishes classification performance, then |
|
34
|
|
|
* small values should be preferred, otherwise, with low number of features, |
|
35
|
|
|
* default value (0.7) will result in satisfactory performance. |
|
36
|
|
|
* |
|
37
|
|
|
* @param mixed $ratio string or float should be given |
|
38
|
|
|
* @return $this |
|
39
|
|
|
* @throws Exception |
|
40
|
|
|
*/ |
|
41
|
|
|
public function setFeatureSubsetRatio($ratio) |
|
42
|
|
|
{ |
|
43
|
|
|
if (is_float($ratio) && ($ratio < 0.1 || $ratio > 1.0)) { |
|
44
|
|
|
throw new Exception("When a float given, feature subset ratio should be between 0.1 and 1.0"); |
|
45
|
|
|
} |
|
46
|
|
|
if (is_string($ratio) && $ratio != 'sqrt' && $ratio != 'log') { |
|
47
|
|
|
throw new Exception("When a string given, feature subset ratio can only be 'sqrt' or 'log' "); |
|
48
|
|
|
} |
|
49
|
|
|
$this->featureSubsetRatio = $ratio; |
|
50
|
|
|
return $this; |
|
51
|
|
|
} |
|
52
|
|
|
|
|
53
|
|
|
/** |
|
54
|
|
|
* @param int $index |
|
55
|
|
|
* @return array |
|
56
|
|
|
*/ |
|
57
|
|
|
protected function getRandomSubset($index) |
|
58
|
|
|
{ |
|
59
|
|
|
list($subset, $targets) = parent::getRandomSubset($index); |
|
60
|
|
|
|
|
61
|
|
|
if (is_float($this->featureSubsetRatio)) { |
|
62
|
|
|
$featureCount = (int)ceil($this->featureSubsetRatio * $this->featureCount); |
|
63
|
|
|
} elseif ($this->featureCount == 'sqrt') { |
|
64
|
|
|
$featureCount = (int)ceil(sqrt($this->featureCount)); |
|
65
|
|
|
} else { |
|
66
|
|
|
$featureCount = (int)log($this->featureCount + 1, 2); |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
if ($featureCount >= $this->featureCount) { |
|
70
|
|
|
$featureCount = $this->featureCount; |
|
71
|
|
|
} |
|
72
|
|
|
|
|
73
|
|
|
$features = range(0, $this->featureCount - 1); |
|
74
|
|
|
shuffle($features); |
|
75
|
|
|
$features = array_slice($features, 0, $featureCount, false); |
|
76
|
|
|
sort($features); |
|
77
|
|
|
$this->classifierColumns[$index] = $features; |
|
78
|
|
|
|
|
79
|
|
|
$columns = []; |
|
80
|
|
|
foreach ($features as $colIndex) { |
|
81
|
|
|
$columns[] = array_column($subset, $colIndex); |
|
82
|
|
|
} |
|
83
|
|
|
$subset= array_map(null, ...$columns); |
|
84
|
|
|
|
|
85
|
|
|
return [$subset, $targets]; |
|
86
|
|
|
} |
|
87
|
|
|
|
|
88
|
|
|
/** |
|
89
|
|
|
* @param array $sample |
|
90
|
|
|
* @return mixed |
|
91
|
|
|
*/ |
|
92
|
|
|
protected function predictSample(array $sample) |
|
93
|
|
|
{ |
|
94
|
|
|
$predictions = []; |
|
95
|
|
|
for ($i=0; $i<count($this->classifiers); $i++) { |
|
|
|
|
|
|
96
|
|
|
$samplePiece = []; |
|
97
|
|
|
foreach ($this->classifierColumns[$i] as $colIndex) { |
|
98
|
|
|
$samplePiece[] = $sample[$colIndex]; |
|
99
|
|
|
} |
|
100
|
|
|
/* @var $classifier Classifier */ |
|
101
|
|
|
$predictions[] = $this->classifiers[$i]->predict($samplePiece); |
|
102
|
|
|
} |
|
103
|
|
|
|
|
104
|
|
|
$counts = array_count_values($predictions); |
|
105
|
|
|
arsort($counts); |
|
106
|
|
|
reset($counts); |
|
107
|
|
|
return key($counts); |
|
108
|
|
|
} |
|
109
|
|
|
} |
|
110
|
|
|
|
php-ai /
php-ml
If the size of the collection does not change during the iteration, it is generally a good practice to compute it beforehand, and not on each iteration: