ML - Lab 7 - kmeans

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import sklearn
import math

# Create Data and scatter them
X = [7, 3, 1, 5, 1, 7, 8, 5]
Y = [1, 4, 5, 8, 3, 8, 2, 9]
labels = ["x" + str(i) for i in range(1, len(X)+1)]
kdata = pd.DataFrame({"X": X, "Y": Y}, index=labels)
plt.figure()
plt.scatter(kdata.X, kdata.Y)
for i in range(len(kdata.index)):
    plt.text(kdata.loc[labels[i], "X"], kdata.loc[labels[i], "Y"], '%s' % (str(labels[i])), size=15, zorder=1)
plt.title("2-D points before clustering")
plt.xlabel("X")
plt.ylabel("Y")
plt.show()


# Create a kmeans model with initial centroids the 3 first points
from sklearn.cluster import KMeans
K = 3
init_centroids = kdata.loc[labels[0:K], :]
kmeans = KMeans(n_clusters=K, init=init_centroids)
kmeans = kmeans.fit(kdata)
print("Final Centroids = ")
print(kmeans.cluster_centers_)
print()
print("Each point's label of centroid = ")
print(kmeans.labels_)
print()


# Cohesion, Separation
print("K = " + str(K))
cohesion = kmeans.inertia_
print("Cohesion = " + str(cohesion))
separation = 0
distance = lambda x1, x2: math.sqrt(((x1.X - x2.X) ** 2) + ((x1.Y - x2.Y) ** 2))
m = kdata.mean()
for i in list(set(kmeans.labels_)):
    mi = kdata.loc[kmeans.labels_ == i, :].mean()
    Ci = len(kdata.loc[kmeans.labels_ == i, :].index)
    separation += Ci * (distance(m, mi) ** 2)
print("Separation = " + str(separation))
print("CSS + BSS = " + str(cohesion + separation))


# Draw centroids
plt.figure()
plt.scatter(kdata.X, kdata.Y, c=kmeans.labels_)
for i in range(len(kdata.index)):
    plt.text(kdata.loc[labels[i], "X"], kdata.loc[labels[i], "Y"], '%s' % (str(labels[i])), size=15, zorder=1)
plt.scatter(kmeans.cluster_centers_[:, 0], kmeans.cluster_centers_[:, 1], marker="+", s=169, c=range(K))
plt.title("After clustering")
plt.xlabel("X")
plt.ylabel("Y")
plt.show()