searching the number of cluster find by AffinityPropagation algo by moving preference value (checking if there is convergence and number of iteration to converge)

looking_best_preference_AffinityPropagation.py

from sklearn.cluster import AffinityPropagation
import pandas as pd
import sys
import cStringIO
# You already have your feature in X

aff_eps = []
for i in [x for x in range(-50, 0, 5)]:
    # To know caputre the output of verbose
    tdout_ = sys.stdout #Keep track of the previous value.
    stream = cStringIO.StringIO()
    sys.stdout = stream
    af = AffinityPropagation(preference=i, max_iter=300, verbose=True).fit(X)
    sys.stdout = sys.__stdout__ # restore the previous stdout.
    output = stream.getvalue()
    if output.split()[0]== "Converged": # converge
        converge = 1
        iteration = float(output.split()[2])
    else: # "Did not converge"
        converge = 0
        iteration = 0
    
    n_clusters_ = len(af.cluster_centers_indices_) 
    print "preference = " +" "+ str(i) +" "+ " cluster = " + str(n_clusters_) +" "+ " Nb iteration " +" "+ str(iteration)
    aff_eps.append({'preference' : i,
                    'n_cluster' : n_clusters_,
                    'verbose': output,
                    'converge' : converge,
                    'iteration' : iteration})
df_aff_eps = pd.DataFrame(aff_eps)
df_aff_eps.set_index('preference', inplace=True)
ax = df_aff_eps.n_cluster.plot(legend=True, x_compat=True)
ax.set_ylabel("Number of Cluster") 
ax = df_aff_eps.iteration.plot(secondary_y=True, style='--', color='r', legend=True)
ax.set_ylabel("Convergence with number of iteration, O = no convergence")