Clustering reverted edits in Wikipedia

Cluster.py

import codecs
import math
import sklearn.cluster
import matplotlib.pyplot as plt

x = set()
c = 0
path = '/home/amir/Downloads/featuresetsforclustering/ptwiki.features_reverted.20k.tsv'
with codecs.open(path, 'r', 'utf-8') as f:
    for line in f:
        line = line.replace('\n', '')
        features = []
        for feature in line.split('\t'):
            if feature == 'False':
                features.append(0)
            elif feature == 'True':
                features.append(1)
            else:
                features.append(float(feature))
        if features[-1] != 1:
            continue
        c += 1
        x.add(tuple(features[:-1]))

print(len(x))
def mean_func(gen):
    mean = 0
    c = 0
    if not gen:
        return 0
    for case in gen:
        c += 1
        mean += case
    return mean/float(c)

def std(gen, mean=None):
    if not gen:
        return 0
    if not mean:
        mean = mean_func(gen)
    variance = 0
    c = 0
    for case in gen:
        c += 1
        variance += (case - mean)**2
    return math.sqrt(variance / float(c))

x_for_scaling = {}
for case in x:
    for i in range(len(case)):
        x_for_scaling[i] = x_for_scaling.get(i, []) + [case[i]]

mean_and_std = {}

for i in x_for_scaling:
    mean = mean_func(x_for_scaling[i])
    std_var = std(x_for_scaling[i], mean)
    mean_and_std[i] = (mean, std_var)

training_set = set()
for case in x:
    new_case = []
    for i in range(len(case)):
        new_case.append((case[i] - mean_and_std[i][0])/mean_and_std[i][1])
    training_set.add(tuple(new_case))

cost_function = {}
res_for_plot = []
for n in range(1, 12):
    classi = sklearn.cluster.KMeans(n_clusters=n)
    res = classi.fit_transform(list(training_set))
    cost_temp = 0
    for case in res:
        cost_temp += min(case)
    cost_function[n] = cost_temp / len(res)
    res_for_plot.append(cost_function[n])
    if n == 2:
        print(classi.cluster_centers_)

print(cost_function)
for i in range(9):
    print i+2,'-', i+1, ':', cost_function[i+2] - cost_function[i+1]
plt.plot(list(range(1, 12)), res_for_plot)
plt.ylabel('Cost function')
plt.xlabel('Number of clusters')
plt.title('Cost function per number of clusters in reverted edits in %s.wp' % path.split('/')[-1][:2])
plt.show()