octoleo/php-ml
1
0
Fork 0
mirror of https://github.com/Llewellynvdm/php-ml.git synced 2025-01-09 16:36:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

Rewrite DBSCAN (#185)

Browse Source 
* Add testcases to DBSCAN

* Fix DBSCAN implementation

* Refactoring DBSCAN implementation

* Fix coding style
This commit is contained in:
Yuji Uchiyama 2018-01-09 18:53:02 +09:00 committed by Arkadiusz Kondas
parent 5a691635d7
commit 9938cf2911
2 changed files with 103 additions and 43 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

112
src/Phpml/Clustering/DBSCAN.php
View File

@ -4,12 +4,13 @@ declare(strict_types=1);
namespace Phpml\Clustering; namespace Phpml\Clustering;
use array_merge;
use Phpml\Math\Distance; use Phpml\Math\Distance;
use Phpml\Math\Distance\Euclidean; use Phpml\Math\Distance\Euclidean;
class DBSCAN implements Clusterer class DBSCAN implements Clusterer
{ {
private const NOISE = -1;
/** /**
* @var float * @var float
*/ */
@ -38,57 +39,82 @@ class DBSCAN implements Clusterer
public function cluster(array $samples): array public function cluster(array $samples): array
{ {
$clusters = []; $labels = [];
$visited = []; $n = 0;
foreach ($samples as $index => $sample) { foreach ($samples as $index => $sample) {
if (isset($visited[$index])) { if (isset($labels[$index])) {
continue; continue;
} }
$visited[$index] = true; $neighborIndices = $this->getIndicesInRegion($sample, $samples);
$regionSamples = $this->getSamplesInRegion($sample, $samples); if (count($neighborIndices) < $this->minSamples) {
if (count($regionSamples) >= $this->minSamples) { $labels[$index] = self::NOISE;
$clusters[] = $this->expandCluster($regionSamples, $visited);
continue;
} }
$labels[$index] = $n;
$this->expandCluster($samples, $neighborIndices, $labels, $n);
++$n;
}
return $this->groupByCluster($samples, $labels, $n);
}
private function expandCluster(array $samples, array $seeds, array &$labels, int $n): void
{
while (($index = array_pop($seeds)) !== null) {
if (isset($labels[$index])) {
if ($labels[$index] === self::NOISE) {
$labels[$index] = $n;
}
continue;
}
$labels[$index] = $n;
$sample = $samples[$index];
$neighborIndices = $this->getIndicesInRegion($sample, $samples);
if (count($neighborIndices) >= $this->minSamples) {
$seeds = array_unique(array_merge($seeds, $neighborIndices));
}
}
}
private function getIndicesInRegion(array $center, array $samples): array
{
$indices = [];
foreach ($samples as $index => $sample) {
if ($this->distanceMetric->distance($center, $sample) < $this->epsilon) {
$indices[] = $index;
}
}
return $indices;
}
private function groupByCluster(array $samples, array $labels, int $n): array
{
$clusters = array_fill(0, $n, []);
foreach ($samples as $index => $sample) {
if ($labels[$index] !== self::NOISE) {
$clusters[$labels[$index]][$index] = $sample;
}
}
// Reindex (i.e. to 0, 1, 2, ...) integer indices for backword compatibility
foreach ($clusters as $index => $cluster) {
$clusters[$index] = array_merge($cluster, []);
} }
return $clusters; return $clusters;
} }
private function getSamplesInRegion(array $localSample, array $samples): array
{
$region = [];
foreach ($samples as $index => $sample) {
if ($this->distanceMetric->distance($localSample, $sample) < $this->epsilon) {
$region[$index] = $sample;
}
}
return $region;
}
private function expandCluster(array $samples, array &$visited): array
{
$cluster = [];
$clusterMerge = [[]];
foreach ($samples as $index => $sample) {
if (!isset($visited[$index])) {
$visited[$index] = true;
$regionSamples = $this->getSamplesInRegion($sample, $samples);
if (count($regionSamples) > $this->minSamples) {
$clusterMerge[] = $regionSamples;
}
}
$cluster[$index] = $sample;
}
$cluster = array_merge($cluster, ...$clusterMerge);
return $cluster;
}
} }

34
tests/Phpml/Clustering/DBSCANTest.php
View File

@ -59,4 +59,38 @@ class DBSCANTest extends TestCase
$this->assertEquals($clustered, $dbscan->cluster($samples)); $this->assertEquals($clustered, $dbscan->cluster($samples));
} }
public function testClusterEpsilonSmall(): void
{
$samples = [[0], [1], [2]];
$clustered = [
];
$dbscan = new DBSCAN($epsilon = 0.5, $minSamples = 2);
$this->assertEquals($clustered, $dbscan->cluster($samples));
}
public function testClusterEpsilonBoundary(): void
{
$samples = [[0], [1], [2]];
$clustered = [
];
$dbscan = new DBSCAN($epsilon = 1.0, $minSamples = 2);
$this->assertEquals($clustered, $dbscan->cluster($samples));
}
public function testClusterEpsilonLarge(): void
{
$samples = [[0], [1], [2]];
$clustered = [
[[0], [1], [2]],
];
$dbscan = new DBSCAN($epsilon = 1.5, $minSamples = 2);
$this->assertEquals($clustered, $dbscan->cluster($samples));
}
} }