06: Učení bez učitele¶

Přednáška:¶

PDF

Cvičení:¶

In [1]:

import numpy as np

from sklearn.cluster import KMeans
from sklearn import datasets

import matplotlib.pyplot as plt

import numpy as np from sklearn.cluster import KMeans from sklearn import datasets import matplotlib.pyplot as plt

Data¶

In [2]:

n_clusters = 5
n_samples = 500

n_clusters = 5 n_samples = 500

In [3]:

X, y_true = datasets.make_blobs(n_samples=n_samples, centers=n_clusters, cluster_std=0.4, random_state=0)
# centers=[[2, 2], [4, 5], [3, 10]]

# data generovana jako kruznice
# X, y_true = datasets.make_circles(n_samples=n_samples, factor=0.5, noise=0.05)

X, y_true = datasets.make_blobs(n_samples=n_samples, centers=n_clusters, cluster_std=0.4, random_state=0) # centers=[[2, 2], [4, 5], [3, 10]] # data generovana jako kruznice # X, y_true = datasets.make_circles(n_samples=n_samples, factor=0.5, noise=0.05)

In [4]:

plt.scatter(X[:, 0], X[:, 1], c=y_true)

plt.scatter(X[:, 0], X[:, 1], c=y_true)

Out[4]:

<matplotlib.collections.PathCollection at 0x73d2b58ca620>

K-Means¶

In [5]:

kmeans = KMeans(n_clusters=n_clusters)

kmeans = KMeans(n_clusters=n_clusters)

In [6]:

kmeans.fit(X)

kmeans.fit(X)

Out[6]:

KMeans(n_clusters=5)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

In [7]:

y_pred = kmeans.predict(X)
plt.scatter(X[:, 0], X[:, 1], c=y_pred)

y_pred = kmeans.predict(X) plt.scatter(X[:, 0], X[:, 1], c=y_pred)

Out[7]:

<matplotlib.collections.PathCollection at 0x73d2b5606bf0>

In [8]:

centers = kmeans.cluster_centers_
print(centers)

centers = kmeans.cluster_centers_ print(centers)

[[ 0.96056297  4.34505773]
 [ 9.28510086 -2.29391247]
 [-1.26927837  7.78360769]
 [ 1.9876471   0.85602104]
 [-1.58354718  2.87334064]]

In [9]:

plt.scatter(X[:, 0], X[:, 1], c=y_pred)
plt.scatter(centers[:, 0], centers[:, 1], c='black', s=200)

plt.scatter(X[:, 0], X[:, 1], c=y_pred) plt.scatter(centers[:, 0], centers[:, 1], c='black', s=200)

Out[9]:

<matplotlib.collections.PathCollection at 0x73d2b5667340>

Kontrola kvality - Analýza siluet¶

In [10]:

from sklearn.metrics import silhouette_samples, silhouette_score
import matplotlib.cm as cm

from sklearn.metrics import silhouette_samples, silhouette_score import matplotlib.cm as cm

In [11]:

silhouette_avg = silhouette_score(X, y_pred)
print(f'Prumerna hodnota analyzy siluet pro {n_clusters} shluku je {silhouette_avg}')

silhouette_avg = silhouette_score(X, y_pred) print(f'Prumerna hodnota analyzy siluet pro {n_clusters} shluku je {silhouette_avg}')

Prumerna hodnota analyzy siluet pro 5 shluku je 0.8126659651729236

In [12]:

silhouette_scores = silhouette_samples(X, y_pred)

silhouette_scores = silhouette_samples(X, y_pred)

In [13]:

np.mean(silhouette_scores)

np.mean(silhouette_scores)

Out[13]:

np.float64(0.8126659651729236)

In [14]:

fig, (ax1, ax2) = plt.subplots(1, 2)

fig.set_figwidth(15)
ax1.set_xlim([-0.2, 1])
ax1.set_xlabel("Koeficient siluet")
ax1.set_ylabel("Shluk")

colour = cm.inferno(y_pred / n_clusters)

y_lower = 10
for i in range(n_clusters):
    silhouette_scores_i = silhouette_scores[y_pred == i]
    silhouette_scores_i.sort()

    y_upper = y_lower + len(silhouette_scores_i)

    ax1.fill_betweenx(
        np.arange(y_lower, y_upper),
        0,
        silhouette_scores_i,
        facecolor=cm.inferno(i / n_clusters)
    )
    ax1.text(-0.1, y_lower + 0.5 * len(silhouette_scores_i), str(i))

    y_lower = y_upper + 10
    

ax1.set_yticks([])
ax1.axvline(silhouette_avg, color='red', linestyle='--')

ax2.scatter(X[:, 0], X[:, 1], c=colour)
ax2.set_xticks([])
ax2.set_yticks([])

fig, (ax1, ax2) = plt.subplots(1, 2) fig.set_figwidth(15) ax1.set_xlim([-0.2, 1]) ax1.set_xlabel("Koeficient siluet") ax1.set_ylabel("Shluk") colour = cm.inferno(y_pred / n_clusters) y_lower = 10 for i in range(n_clusters): silhouette_scores_i = silhouette_scores[y_pred == i] silhouette_scores_i.sort() y_upper = y_lower + len(silhouette_scores_i) ax1.fill_betweenx( np.arange(y_lower, y_upper), 0, silhouette_scores_i, facecolor=cm.inferno(i / n_clusters) ) ax1.text(-0.1, y_lower + 0.5 * len(silhouette_scores_i), str(i)) y_lower = y_upper + 10 ax1.set_yticks([]) ax1.axvline(silhouette_avg, color='red', linestyle='--') ax2.scatter(X[:, 0], X[:, 1], c=colour) ax2.set_xticks([]) ax2.set_yticks([])

Out[14]:

[]

In [ ]: