added k-means clustering

Author: NotShrirang

K-Means Clustering/KMeans.py

@@ -0,0 +1,70 @@
+import numpy as np
+
+
+class K_Means:
+    def __init__(self, k: int = 2, tolerence = 0.001, max_itr = 300):
+        """Class for K-Means Clustering
+
+        Args:
+            - k (int, optional): total cluster numbers. Defaults to 2.
+            - tolerence (float, optional): tolerence to centroids movement. Defaults to 0.001.
+            - max_itr (int, optional): maximum iterations till which clustering is done. Defaults to 300.
+        """
+        self.k = k
+        self.tol = tolerence
+        self.max_itr = max_itr
+
+    def fit(self, X, y = None) -> list:
+        """Fit method for training the model.
+
+        Args:
+            - X (arrayLike, list, npArray): data for training.
+            - y (arrayLike, list, npArray, optional): For consistency of API. Even if inserted, it is ignored. Defaults to None.
+        """
+        self.centroids = {}
+        self.__moves_history = []
+        for i in range(self.k):
+            self.centroids[i] = X[i]
+
+        for i in range(self.max_itr):
+            self.classifications = {}
+
+            for i in range(self.k):
+                self.classifications[i] = []
+
+            for featureset in X:
+                distances = [np.linalg.norm(featureset-self.centroids[centroid]) for centroid in self.centroids]
+                classification = distances.index(min(distances))
+                self.classifications[classification].append(featureset)
+
+            prev_centroids = dict(self.centroids)
+            for classification in self.classifications:
+                self.centroids[classification] = np.average(self.classifications[classification], axis=0)
+
+            optimized = True
+
+            for c in self.centroids:
+                original_centroid = prev_centroids[c]
+                current_centroid  = self.centroids[c]
+                move = np.sum((current_centroid-original_centroid) / original_centroid*100.0)
+                if move > self.tol:
+                    self.__moves_history.append(move)
+                    optimized = False
+
+            if optimized:
+                break
+        
+        return self.__moves_history
+
+    def predict(self, X):
+        """Predict method for predicting values.
+
+        Args:
+            - X (arrayLike, list, npArray): input data.
+        """
+        predicted = []
+        for point in X:
+            distances = [np.linalg.norm(point-self.centroids[centroid]) for centroid in self.centroids]
+            classification = distances.index(min(distances))
+            predicted.append(classification)
+        return predicted