Das deutsche Python-Forum

Code: Alles auswählen

#!/usr/bin/env python
# coding: utf-8

# In[6]:


import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import make_blobs


# In[12]:


data = make_blobs(n_samples=200, n_features=2, 
                           centers=4, cluster_std=1.8,random_state=101)

plt.scatter(data[0][:,0],data[0][:,1],c=data[1],cmap='rainbow')
feature,cluster = data


# In[30]:



indices = np.random.choice(len(feature),k)
initial = np.copy(feature[initialmeans])
point_classification = np.full(len(feature),-1)



# In[57]:


def kmeans(data, k, centroids):
    alldistances = np.array([np.linalg.norm(data - centroid, axis=1)
                             for centroid in centroids])
    assert alldistances.shape == (k, len(data))
    point_classification = np.argmin(alldistances, axis=0)
    new_centroids = np.array([np.mean(data[point_classification == pointclass],
                                      axis=0)
                              for pointclass in np.arange(k)])
    return point_classification, new_centroids
    
point_classification, new_centroids = kmeans(feature, 4, initial)   


# Ergebnisse betrachten:

fig, ax = plt.subplots(figsize=(10,10))
ax.set_xlim(np.min(feature[:,0])-0.2, np.max(feature[:,0])+0.2)

Code: Alles auswählen

#!/usr/bin/env python
# coding: utf-8

# In[1]:


import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import make_blobs


# In[18]:


data = make_blobs(n_samples=200, n_features=2, 
                           centers=4, cluster_std=1.8,random_state=101)

plt.scatter(data[0][:,0],data[0][:,1],c=data[1],cmap='rainbow')
feature,cluster = data


# In[19]:


k = 4
random_indices = np.random.choice(len(feature), k)
initial_cs = feature[random_indices]
cs = np.copy(initial_cs)


# In[20]:


def kmeans(data, k, centroids):
    alldistances = np.array([np.linalg.norm(data - centroid, axis=1)
                             for centroid in centroids])
    assert alldistances.shape == (k, len(data))
    point_classification = np.argmin(alldistances, axis=0)
    new_centroids = np.array([np.mean(data[point_classification == pointclass],
                                      axis=0)
                              for pointclass in np.arange(k)])
    return point_classification, new_centroids
    
point_classification, new_centroids = kmeans(feature, 4, initial)   


# Ergebnisse betrachten:

fig, ax = plt.subplots(figsize=(10,10))
ax.set_xlim(np.min(feature[:,0])-0.2, np.max(feature[:,0])+0.2)
ax.set_ylim(np.min(feature[:,1])-0.2, np.max(feature[:,1])+0.2)
ax.scatter(feature[:,0], feature[:,1] , c =point_classification)
ax.plot(new_centroids[:,0],new_centroids[:,1],'+', c = 'black',markersize = 20)