public function train()
{
$svm = new SVM();
$svm->setOptions(array(SVM::OPT_KERNEL_TYPE => SVM::KERNEL_LINEAR));
$trainings = CM_Db_Db::select('cm_svmtraining', array('class', 'values'), array('svmId' => $this->getId()))->fetchAll();
$problem = array();
$classCounts = array();
foreach ($trainings as $training) {
$class = $training['class'];
$values = unserialize($training['values']);
if (!isset($classCounts[$class])) {
$classCounts[$class] = 0;
}
$classCounts[$class]++;
$problem[] = array_merge(array(0 => $class), $values);
}
$weights = array();
foreach ($classCounts as $class => $count) {
$weights[$class] = min($classCounts) / $count;
}
if (empty($weights)) {
$weights = null;
}
$this->_model = $svm->train($problem, $weights);
$this->_model->save($this->_getFilePath());
}
$minus = array();
foreach ($data as $el) {
if ($el[0] == 1) {
$plus[] = $el;
} else {
$minus[] = $el;
}
}
$SVMProblems = array();
for ($i = 0; $i < 10; $i++) {
shuffle($plus);
shuffle($minus);
$problem = array_merge(array_slice($plus, 0, count($plus) / 2), array_slice($minus, count($minus) / 2));
$testProblem = array_merge(array_slice($plus, count($plus) / 2), array_slice($minus, 0, count($minus) / 2));
// SVM object
$svm = new SVM();
$svm->setOptions(array(SVM::OPT_TYPE => SVM::C_SVC, SVM::OPT_KERNEL_TYPE => SVM::KERNEL_LINEAR));
$model = $svm->train($problem);
$target = array();
foreach ($testProblem as $p) {
unset($p[0]);
$target[] = $model->predict($p);
}
$SVMProblems[] = new SVMProblem($testProblem, $target);
}
?>
<!DOCTYPE html>
<html lang="">
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<?php
require "SVM.php";
$data = array(array(-1, 1 => 0.43, 3 => 0.12, 9284 => 0.2), array(1, 1 => 0.22, 5 => 0.01, 94 => 0.11));
$svm = new SVM();
$model = $svm->train($data);
$data = array(1 => 0.43, 3 => 0.12, 9284 => 0.2);
$result = $model->predict($data);
var_dump($result);
$model->save('model.svm');
