def plot_views_cits_corr():
input_cits = open('data/series_cits.json', 'r')
series_cits = json.load(input_cits)
months_cits = series_cits['months']
months_cits = [
int(m.split('-')[0]) + int(m.split('-')[1]) / 12 for m in months_cits
]
data = read_file.load_data()
data = read_file.filter_outliers(data)
corrs = []
for i, _, _, xs, ys, _ in data:
try:
idx_begin = months_cits.index(xs[0])
idx_end = months_cits.index(xs[-1])
x = series_cits['data'][i]['citations'][idx_begin:idx_end + 1]
x = np.diff(x)
y = np.diff(ys)
if np.count_nonzero(x[0] == x) == len(x) or np.count_nonzero(
y[0] == y) == len(y):
continue
if np.count_nonzero(x) > len(x) / 2 and np.count_nonzero(
y) > len(y) / 2:
corr = pearsonr(x, y)[0]
corrs.append(corr)
except:
pass
plt.hist(corrs, bins=100)
plt.title('correlação entre número de visualizações e número de citações')
plt.savefig('corr_views_cits.pdf')
def get_dois(n, filename):
data = read_file.load_data(filename)
data = read_file.filter_outliers(data)
dois = []
for i, s, b, xs, ys, p in data:
if len(s) == n:
dois.append(i)
return dois
def get_data_by_number_segm():
data = read_file.load_data()
data = read_file.filter_outliers(data)
freq_delta_t = defaultdict(lambda: [])
freq_views = defaultdict(lambda: [])
for sample in data:
N = len(sample[1])
delta_t = sample[3][-1] - sample[3][0]
views = sample[-2][-1]
freq_delta_t[N].append(delta_t)
freq_views[N].append(views)
return freq_delta_t, freq_views
def get_all_data():
data = read_file.load_data()
data = read_file.filter_outliers(data)
# freq1 = defaultdict(lambda:[])
freq = []
for sample in data:
N = len(sample[1])
delta_t = sample[3][-1] - sample[3][0]
# freq1[N].append((delta_t,sample[-2][-1]))
views = sample[-2][-1]
if views == 0:
continue
freq.append((delta_t, views))
return freq
def plt_views(dois, labels):
data = read_file.load_data()
data = read_file.filter_outliers(data)
views = []
for i, s, b, xs, ys, p in data:
if i in dois:
views.append(ys[-1])
views = np.asarray(views)
values = []
unique = np.unique(labels)
fig, axes = plt.subplots(1, 3, sharex=True, sharey=True, figsize=(9, 3))
for u in unique:
if u == -1:
continue
axes[u].hist(views[labels == u], label=u, bins=30)
axes[u].legend()
plt.savefig('views_dist_labels.pdf')
import json
import numpy as np
from read_file import load_data, filter_outliers
sources = ['clusters\\clusters\\clusters_ind_single_0.50_2.txt',
'clusters\\clusters\\clusters_ind_single_0.35_3.txt',
'clusters\\clusters\\clusters_ind_single_0.47_4.txt',
'clusters\\clusters\\clusters_ind_single_0.56_5.txt']
data = load_data()
data = filter_outliers(data)
for N, source in zip([2, 3, 4, 5], sources):
labels = np.loadtxt(source, dtype=np.int).tolist()
unique, counts = np.unique(labels, return_counts=True)
unique = unique[counts >= 10]
counts = counts[counts >= 10]
unique_idxs = np.argsort(counts)[-3:]
unique = unique[unique_idxs].tolist()
labels = [unique.index(l) if l in unique else -1 for l in labels]
dois = []
for i, s, b, xs, ys, p in data:
if len(s) == N:
dois.append(i)
print(len(dois), len(labels))
doi2cluster = dict(zip(dois, labels))
str_json = json.dumps(doi2cluster)
out = open('doi2cluster_%d_3.json' % N, 'w')
out.write(str_json)
out.close()
sources = [
'clusters\\clusters\\clusters_ind_single_0.50_2.txt',
'clusters\\clusters\\clusters_ind_single_0.35_3.txt',
'clusters\\clusters\\clusters_ind_single_0.47_4.txt',
'clusters\\clusters\\clusters_ind_single_0.56_5.txt'
]
labels3 = np.loadtxt(sources[1], dtype=np.int)
unique, count = np.unique(labels3, return_counts=True)
unique = unique[count >= 10]
count = count[count >= 10]
unique_idxs = np.argsort(count)[-3:]
unique = unique[unique_idxs].tolist()
labels3 = [unique.index(l) if l in unique else -1 for l in labels3]
data = read_file.load_data()
data = read_file.filter_outliers(data)
# dois = {2: [], 3: [], 4: [], 5: []}
# slopes = []
# intervals = []
# for i, s, b, xs, ys, p in data:
# dois[len(s)].append(i)
#
# print('1y')
# plt_cits(dois[3], labels3, 1)
# print('2y')
# plt_cits(dois[3], labels3, 2)
# print('3y')
# plt_cits(dois[3], labels3, 5)
# # pegar os labels
# # ver a distribuição dos grupos kk views cits tweets