1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
declare(strict_types=1); |
4
|
|
|
|
5
|
|
|
namespace Phpml\Classification\Ensemble; |
6
|
|
|
|
7
|
|
|
use Phpml\Helper\Predictable; |
8
|
|
|
use Phpml\Helper\Trainable; |
9
|
|
|
use Phpml\Math\Statistic\Mean; |
10
|
|
|
use Phpml\Classification\Classifier; |
11
|
|
|
use Phpml\Classification\DecisionTree; |
12
|
|
|
use Phpml\Classification\NaiveBayes; |
13
|
|
|
|
14
|
|
|
class Bagging implements Classifier |
15
|
|
|
{ |
16
|
|
|
use Trainable, Predictable; |
17
|
|
|
|
18
|
|
|
/** |
19
|
|
|
* @var int |
20
|
|
|
*/ |
21
|
|
|
protected $numSamples; |
22
|
|
|
|
23
|
|
|
/** |
24
|
|
|
* @var array |
25
|
|
|
*/ |
26
|
|
|
private $targets = []; |
27
|
|
|
|
28
|
|
|
/** |
29
|
|
|
* @var int |
30
|
|
|
*/ |
31
|
|
|
protected $featureCount = 0; |
32
|
|
|
|
33
|
|
|
/** |
34
|
|
|
* @var int |
35
|
|
|
*/ |
36
|
|
|
protected $numClassifier; |
37
|
|
|
|
38
|
|
|
/** |
39
|
|
|
* @var Classifier |
40
|
|
|
*/ |
41
|
|
|
protected $classifier = DecisionTree::class; |
42
|
|
|
|
43
|
|
|
/** |
44
|
|
|
* @var array |
45
|
|
|
*/ |
46
|
|
|
protected $classifierOptions = ['depth' => 20]; |
47
|
|
|
|
48
|
|
|
/** |
49
|
|
|
* @var array |
50
|
|
|
*/ |
51
|
|
|
protected $classifiers; |
52
|
|
|
|
53
|
|
|
/** |
54
|
|
|
* @var float |
55
|
|
|
*/ |
56
|
|
|
protected $subsetRatio = 0.5; |
57
|
|
|
|
58
|
|
|
/** |
59
|
|
|
* @var array |
60
|
|
|
*/ |
61
|
|
|
private $samples = []; |
62
|
|
|
|
63
|
|
|
/** |
64
|
|
|
* Creates an ensemble classifier with given number of base classifiers<br> |
65
|
|
|
* Default number of base classifiers is 100. |
66
|
|
|
* The more number of base classifiers, the better performance but at the cost of procesing time |
67
|
|
|
* |
68
|
|
|
* @param int $numClassifier |
69
|
|
|
*/ |
70
|
|
|
public function __construct($numClassifier = 50) |
71
|
|
|
{ |
72
|
|
|
$this->numClassifier = $numClassifier; |
73
|
|
|
} |
74
|
|
|
|
75
|
|
|
/** |
76
|
|
|
* This method determines the ratio of samples used to create the 'bootstrap' subset, |
77
|
|
|
* e.g., random samples drawn from the original dataset with replacement (allow repeats), |
78
|
|
|
* to train each base classifier. |
79
|
|
|
* |
80
|
|
|
* @param float $ratio |
81
|
|
|
* @return $this |
82
|
|
|
* @throws Exception |
83
|
|
|
*/ |
84
|
|
|
public function setSubsetRatio(float $ratio) |
85
|
|
|
{ |
86
|
|
|
if ($ratio < 0.1 || $ratio > 1.0) { |
87
|
|
|
throw new \Exception("Subset ratio should be between 0.1 and 1.0"); |
88
|
|
|
} |
89
|
|
|
$this->subsetRatio = $ratio; |
90
|
|
|
return $this; |
91
|
|
|
} |
92
|
|
|
|
93
|
|
|
/** |
94
|
|
|
* This method is used to set the base classifier. Default value is |
95
|
|
|
* DecisionTree::class, but any class that implements the <i>Classifier</i> |
96
|
|
|
* can be used. <br> |
97
|
|
|
* While giving the parameters of the classifier, the values should be |
98
|
|
|
* given in the order they are in the constructor of the classifier and parameter |
99
|
|
|
* names are neglected. |
100
|
|
|
* |
101
|
|
|
* @param string $classifier |
102
|
|
|
* @param array $classifierOptions |
103
|
|
|
* @return $this |
104
|
|
|
*/ |
105
|
|
|
public function setClassifer(string $classifier, array $classifierOptions = []) |
106
|
|
|
{ |
107
|
|
|
$this->classifier = $classifier; |
|
|
|
|
108
|
|
|
$this->classifierOptions = $classifierOptions; |
109
|
|
|
return $this; |
110
|
|
|
} |
111
|
|
|
|
112
|
|
|
/** |
113
|
|
|
* @param array $samples |
114
|
|
|
* @param array $targets |
115
|
|
|
*/ |
116
|
|
|
public function train(array $samples, array $targets) |
117
|
|
|
{ |
118
|
|
|
$this->samples = array_merge($this->samples, $samples); |
119
|
|
|
$this->targets = array_merge($this->targets, $targets); |
120
|
|
|
$this->featureCount = count($samples[0]); |
121
|
|
|
$this->numSamples = count($this->samples); |
122
|
|
|
|
123
|
|
|
// Init classifiers and train them with random sub-samples |
124
|
|
|
$this->classifiers = $this->initClassifiers(); |
125
|
|
|
$index = 0; |
126
|
|
|
foreach ($this->classifiers as $classifier) { |
127
|
|
|
list($samples, $targets) = $this->getRandomSubset($index); |
128
|
|
|
$classifier->train($samples, $targets); |
129
|
|
|
++$index; |
130
|
|
|
} |
131
|
|
|
} |
132
|
|
|
|
133
|
|
|
/** |
134
|
|
|
* @param int $index |
135
|
|
|
* @return array |
136
|
|
|
*/ |
137
|
|
|
protected function getRandomSubset($index) |
138
|
|
|
{ |
139
|
|
|
$subsetLength = (int)ceil(sqrt($this->numSamples)); |
|
|
|
|
140
|
|
|
$denom = $this->subsetRatio / 2; |
141
|
|
|
$subsetLength = $this->numSamples / (1 / $denom); |
142
|
|
|
$index = $index * $subsetLength % $this->numSamples; |
143
|
|
|
$samples = []; |
144
|
|
|
$targets = []; |
145
|
|
|
for ($i=0; $i<$subsetLength * 2; $i++) { |
146
|
|
|
$rand = rand($index, $this->numSamples - 1); |
147
|
|
|
$samples[] = $this->samples[$rand]; |
148
|
|
|
$targets[] = $this->targets[$rand]; |
149
|
|
|
} |
150
|
|
|
return [$samples, $targets]; |
151
|
|
|
} |
152
|
|
|
|
153
|
|
|
/** |
154
|
|
|
* @return array |
155
|
|
|
*/ |
156
|
|
|
protected function initClassifiers() |
157
|
|
|
{ |
158
|
|
|
$classifiers = []; |
159
|
|
|
for ($i=0; $i<$this->numClassifier; $i++) { |
160
|
|
|
$ref = new \ReflectionClass($this->classifier); |
161
|
|
|
if ($this->classifierOptions) { |
|
|
|
|
162
|
|
|
$obj = $ref->newInstanceArgs($this->classifierOptions); |
163
|
|
|
} else { |
164
|
|
|
$obj = $ref->newInstance(); |
165
|
|
|
} |
166
|
|
|
$classifiers[] = $this->initSingleClassifier($obj, $i); |
167
|
|
|
} |
168
|
|
|
return $classifiers; |
169
|
|
|
} |
170
|
|
|
|
171
|
|
|
/** |
172
|
|
|
* @param Classifier $classifier |
173
|
|
|
* @param int $index |
174
|
|
|
* @return Classifier |
175
|
|
|
*/ |
176
|
|
|
protected function initSingleClassifier($classifier, $index) |
|
|
|
|
177
|
|
|
{ |
178
|
|
|
return $classifier; |
179
|
|
|
} |
180
|
|
|
|
181
|
|
|
/** |
182
|
|
|
* @param array $sample |
183
|
|
|
* @return mixed |
184
|
|
|
*/ |
185
|
|
|
protected function predictSample(array $sample) |
186
|
|
|
{ |
187
|
|
|
$predictions = []; |
188
|
|
|
foreach ($this->classifiers as $classifier) { |
189
|
|
|
/* @var $classifier Classifier */ |
190
|
|
|
$predictions[] = $classifier->predict($sample); |
191
|
|
|
} |
192
|
|
|
|
193
|
|
|
$counts = array_count_values($predictions); |
194
|
|
|
arsort($counts); |
195
|
|
|
reset($counts); |
196
|
|
|
return key($counts); |
197
|
|
|
} |
198
|
|
|
} |
199
|
|
|
|
Our type inference engine has found an assignment to a property that is incompatible with the declared type of that property.
Either this assignment is in error or the assigned type should be added to the documentation/type hint for that property..