Perform data clustering using any distance or centroid metric as euclidian or cosine distance and median or mean centroid

k_custom.py

# adapted from https://pythonprogramming.net/k-means-from-scratch-machine-learning-tutorial/

from sklearn import datasets
import numpy as np
from scipy import spatial

euclidian_distance = lambda a, b: np.linalg.norm(a - b)
cosine_distance = lambda a, b: spatial.distance.cosine(a, b)

class KCustom:
    def __init__(self, k=2, tol=0.001, max_iter=300, centroid_func=np.average, distance_func=euclidian_distance):
        self.k = k
        self.tol = tol
        self.max_iter = max_iter
        self.centroid_fun = centroid_func
        self.distance_func = distance_func

    def fit(self,data):

        self.centroids = {}

        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:
                distances = [self.distance_func(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] = self.centroid_fun(self.classifications[classification],axis=0)

        self.bsse = {}
        self.wsse = 0.0

        for k, v in self.classifications.items():
            v = np.vstack(v)
            bsse = np.sum(np.var(v) * len(v) - 1)
            self.wsse += bsse
            self.bsse[k] = bsse

# test distance methods
a = np.array([0.2,0.4,0.4])
b = np.array([0.2,0.3,0.4])
cos_sim = cosine_distance(a, b)
eucl_sim = euclidian_distance(a, b)
print(cos_sim, eucl_sim)

# cluster iris data set
iris = datasets.load_iris()
#model = KCustom(k=10, centroid_func=np.average, distance_func=euclidian_distance)
#model = KCustom(k=25, centroid_func=np.median, distance_func=cosine_distance)
model = KCustom(k=25, centroid_func=np.median, distance_func=euclidian_distance)
model.fit(iris.data)

# Within Set Sum of Squared Errors (similar to sklearn.cluster.KMeans#inertia_)
model.wsse

# Center points (similar to sklearn.cluster.KMeans#cluster_centers_)
model.centroids

# Cluster dict (key: label, value: data points)
model.classifications