k-means clustering K平均算法

此算法的主要作用：屏幕上很多的点，把相邻的点聚到离他最近的点。

k-means algorithm算法是一个聚类算法，把n个对象根据他们的属性分为k个分割，k < n。它与处理混合正态分布的最大期望算法很相似，因为他们都试图找到数据中自然聚类的中心。

php实现算法代码如下：

</pre><pre name="code" class="php"></pre><pre name="code" class="php">class Cluster { public $points; public $avgPoint; function calculateAverage($maxX, $maxY) { if (count($this->points)==0) { $this->avgPoint->x = rand(0, $maxX); $this->avgPoint->y = rand(0,$maxY); //we didn't get any clues at all :( lets just randomize and hope for better... return; } foreach($this->points as $p) { $xsum += $p->x; $ysum += $p->y; } $count = count($this->points); $this->avgPoint->x = $xsum / $count; $this->avgPoint->y = $ysum / $count; } } class Point { public $x; public $y; function getDistance($p) { $x1 = $this->x - $p->x; $y1 = $this->y - $p->y; return sqrt($x1*$x1 + $y1*$y1); } } function distributeOverClusters($k, $arr) { foreach($arr as $p) { if ($p->x > $maxX) $maxX = $p->x; if ($p->y > $maxY) $maxY = $p->y; } $clusters = array(); for($i = 0; $i < $k; $i++) { $clusters[] = new Cluster(); $tmpP = new Point(); $tmpP->x=rand(0,$maxX); $tmpP->y=rand(0,$maxY); $clusters[$i]->avgPoint = $tmpP; } #deploy points to closest center. #recalculate centers for ($a = 0; $a < 200; $a++) # run it 200 times { foreach($clusters as $cluster) $cluster->points = array(); //reinitialize foreach($arr as $pnt) { $bestcluster=$clusters[0]; $bestdist = $clusters[0]->avgPoint->getDistance($pnt); foreach($clusters as $cluster) { if ($cluster->avgPoint->getDistance($pnt) < $bestdist) { $bestcluster = $cluster; $bestdist = $cluster->avgPoint->getDistance($pnt); } } $bestcluster->points[] = $pnt;//add the point to the best cluster. } //recalculate the centers. foreach($clusters as $cluster) $cluster->calculateAverage($maxX, $maxY); } return $clusters; } $p = new Point(); $p->x = 2; $p->y = 2; $p2 = new Point(); $p2->x = 3; $p2->y = 2; $p3 = new Point(); $p3->x = 8; $p3->y = 2; $arr[] = $p; $arr[] = $p2; $arr[] = $p3; var_dump(distributeOverClusters(2, $arr));

参考地址：http://en.wikipedia.org/wiki/K-means_clustering