Merge pull request deutranium#67 from blackfly19/master

KMeans in python

Author: deutranium

MachineLearningAlgorithms/KMeans/KMeans.py

@@ -0,0 +1,64 @@
+import random
+import itertools
+import numpy as np
+from sklearn.datasets import make_blobs
+
+class KMeans:
+	def __init__(self,k):
+		self.k = k
+		self.means = None
+	
+	def form_clusters(self,data,indexes):
+		#assigning clusters to points
+
+		clusters = [[] for i in range(self.k)]
+		means = []
+		for point,index in zip(data,indexes):
+			clusters[index].append(point)
+		for cluster in clusters:
+			if cluster:
+				means.append(np.mean(cluster,axis=0).tolist())
+			else:
+				means.append(random.choice(data))
+		return means	
+
+	def classify(self,vector):
+		
+		diff = []
+		#calculating euclidean distance for each point from each center
+		for mean in self.means:
+			diff.append(np.linalg.norm(mean - vector))
+		return np.argmin(diff)
+
+	@property
+	def centers(self):
+		return self.means	
+				
+	def fit(self,data):
+
+		#random initialization of means
+		cluster_index = [random.randrange(self.k) for _ in data]
+		for i in itertools.count():
+			#computing means and reassigning centers
+			self.means = self.form_clusters(data,cluster_index)
+			new_cluster_index = [self.classify(val) for val in data]
+			
+			#if difference between old centers and new centers is 0 then break the loop
+			differences = 0
+			for ind,new_ind in zip(cluster_index,new_cluster_index):
+				if ind != new_ind:
+					differences = differences + 1
+
+			if differences == 0:
+			 	return
+
+			#calculating means for new cluster centers
+			cluster_index = new_cluster_index
+			self.means = self.form_clusters(data,cluster_index)
+
+#generating random dataset
+X,y= make_blobs(n_samples=150,n_features=2,centers=3)
+clustering = KMeans(k=3)
+clustering.fit(X)
+print(clustering.centers)
+