Building Clusters as Hierarchical Trees

xxxxxxxxxx

 

1

# suppress ClusterWarning

2

import warnings; warnings.simplefilter('ignore')

3

import sklearn.cluster as scl

4

import numpy as np,pandas as pd,pylab as pl

5

from IPython.core.magic import register_line_magic

6

from scipy.spatial.distance import pdist,squareform

7

from scipy.cluster.hierarchy import linkage,dendrogram

8

variables=['a','b','c']

9

labels=['id'+str(i) for i in range(1,10)]

10

X=np.random.random_sample([9,3])*10

11

df=pd.DataFrame(X,columns=variables,index=labels)

12

ac=scl.AgglomerativeClustering(

13

    n_clusters=3,affinity='euclidean',linkage='complete')

14

ac_labels=ac.fit_predict(X)

15

print('Cluster labels: %s'%ac_labels)

16

@register_line_magic

17

def dist_clusters(method):

18

    row_dist=pdist(df,metric='euclidean')

19

    if method=='False':

20

        row_dist=squareform(row_dist)

21

    row_clusters=linkage(

22

        row_dist,method='complete',metric='euclidean')

23

    c=row_clusters.shape[0]

24

    rc=pd.DataFrame(

25

        row_clusters,

26

        columns=['row label 1','row label 2','distance','n_items'],

27

        index=['cl%d'%(i+1) for i in range(c)])

28

    display(rc)

29

    fig=pl.figure(figsize=(6.5,5))

30

    axd=fig.add_axes([.1,.1,.31,.6])

31

    d=dendrogram(row_clusters,orientation='left',labels=labels)

32

    df_cl=df.iloc[d['leaves'][::-1]]

33

    axd.set_xticks([]); axd.set_yticks([])

34

    for i in axd.spines.values(): i.set_visible(False)

35

    axm=fig.add_axes([.2,.1,.6,.6])

36

    cax=axm.matshow(df_cl,interpolation='nearest',cmap='cool')

37

    fig.colorbar(cax)

38

    ac=[str(ac_labels[int(el[2])-1])+'-'+el 

39

        for el in list(df_cl.index)]

40

    axm.set_xticks(axm.get_xticks().tolist())

41

    axm.set_yticks(axm.get_yticks().tolist())

42

    axm.set_xticklabels(['']+list(df_cl.columns)+[''])

43

    axm.set_yticklabels(['']+ac+[''])

44

    if method=='False': pl.title('incorrect')

45

    else: pl.title('correct')

46

    pl.show()

47

%dist_clusters False

48

%dist_clusters True

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages