Create kMeans.py

ML_Algorithms/K-Means/kMeans.py

@@ -0,0 +1,65 @@
+# Code for kMeans algorithm which is an Unsupervised Learning algorithm, which groups the unlabeled dataset into different clusters
+
+import numpy as np
+
+class K_Means:
+    def __init__(self, k=2, max_iter=100, tol=0.001):  
+        self.k = k
+        # k= number of clusters
+        self.max_iter = max_iter
+        #  max_iter= maximum number of times to iterate through the dataset 
+        self.tol = tol
+        # tol= maximum tolerance level in the output
+
+
+    def fit(self,data):
+        # Initializing the centroid for each cluster 
+        self.centroids = {}
+
+        # Taking random points as centroid
+        for i in range(self.k):
+            self.centroids[i] = data[i]
+
+        for i in range(self.max_iter):
+            self.classifications = {}
+
+            for i in range(self.k):
+                self.classifications[i] = []
+
+            for featureset in data:
+
+                # Calculating distance from each centroid
+                distances = [np.linalg.norm(featureset-self.centroids[centroid]) for centroid in self.centroids]
+
+                #Finding the centroid from which distance is minimum
+                classification = distances.index(min(distances))
+                self.classifications[classification].append(featureset)
+
+            prev_centroids = dict(self.centroids)
+
+            # Finding the new centroid on the basis of new classified points
+            for classification in self.classifications:
+                self.centroids[classification] = np.average(self.classifications[classification],axis=0)
+
+            optimized = True
+
+            #Checking if the centroid each cluster converge
+            for c in self.centroids:
+                original_centroid = prev_centroids[c]
+                current_centroid = self.centroids[c]
+
+                # Comparing the previous and new centroids
+                if np.sum((current_centroid-original_centroid)/original_centroid*100.0) > self.tol:
+                    # print(np.sum((current_centroid-original_centroid)/original_centroid*100.0))
+                    optimized = False
+
+            # If the centroids converge then it is done else again iterate through the whole dataset and find the new centroids   
+            if optimized:
+                break
+
+    def predict(self,data):
+        # Calculating the distance of test data from the centroid of each cluster
+        distances = [np.linalg.norm(data-self.centroids[centroid]) for centroid in self.centroids]
+        # Classify the data with the cluster from which distance is minimum 
+        classification = distances.index(min(distances))
+        return classification