def extract_features(eid, tree, label):
features = {}
# 初始化
for i in range(1, 11):
features[i] = {
'count': 0,
# '社交网络特征'
'rep_count': 0,
'comments_count': 0,
# 文本特征
# 'text_length': 0,
# 'text_NN_rat': 0,
# 'text_verb_rat': 0,
# 'text_adj_rat': 0,
#
# 'pos_count': 0,
# 'neg_count': 0,
# 'neu_count': 0,
# '@_count': 0,
# 'stopword_count': 0,
# 用户特征
'bi_followers_count': 0,
# 'user_des_len': 0,
'friends_count': 0,
'verified_count': 0,
'followers_count': 0,
'statuses_count': 0,
'male_count': 0, # m男 f女
'female_count': 0,
'favourites_count': 0,
}
#提取特征，从第一层到第十层
for node in tree.all_nodes_itr():
level = tree.depth(node=node)
if level <= 10 and level > 0: #只统计1到10层的个数
features[level]['count'] += 1
features[level]['rep_count'] += node.data.reposts_count
features[level]['comments_count'] += node.data.comments_count
# features[level]['pos_count'] += node.data.reposts_count
# features[level]['neg_count'] += node.data.reposts_count
# features[level]['neu_count'] += node.data.reposts_count
features[level][
'bi_followers_count'] += node.data.bi_followers_count
features[level]['friends_count'] += node.data.friends_count
if node.data.verified == True:
features[level]['verified_count'] += node.data.verified
features[level]['followers_count'] += node.data.followers_count
features[level]['statuses_count'] += node.data.statuses_count
if node.data.gender == 'm':
features[level]['male_count'] += 1
elif node.data.gender == 'f':
features[level]['female_count'] += 1
features[level]['favourites_count'] += node.data.favourites_count
#对第一层到第十层的特征求平均值
for i in range(1, 11):
if features[i]['count'] != 0:
features[i].update({
'rep_count':
round(features[i]['rep_count'] / features[i]['count'], 2),
'comments_count':
round(features[i]['comments_count'] / features[i]['count'], 2),
'bi_followers_count':
round(features[i]['bi_followers_count'] / features[i]['count'],
2),
'friends_count':
round(features[i]['friends_count'] / features[i]['count'], 2),
'verified_count':
round(features[i]['verified_count'] / features[i]['count'], 2),
'followers_count':
round(features[i]['followers_count'] / features[i]['count'],
2),
'statuses_count':
round(features[i]['statuses_count'] / features[i]['count'], 2),
'male_count':
round(features[i]['male_count'] / features[i]['count'], 2),
'female_count':
round(features[i]['female_count'] / features[i]['count'], 2),
'favourites_count':
round(features[i]['favourites_count'] / features[i]['count'],
2),
})
return features
def cal_node_level_count(type):
data = pd.read_csv('D:/chenjiao/SinaWeibo/datasets2/Weibo.txt',
sep='\t',
header=None)
if type == 'fake':
data = data.loc[data[1] == 'label:1']
elif type == 'real':
data = data.loc[data[1] == 'label:0']
data_array = data.as_matrix()
tree_dict_list = []
max_depth = 0
all_infos = []
for i in range(data_array.shape[0]):
eid = str(data_array[i][0]).replace('eid:', '')
label = str(data_array[i][1].replace('label:', ''))
load_f = open(
'D:/chenjiao/SinaWeibo/datasets2/Weibo/{}.json'.format(eid),
'r',
encoding='utf-8')
json_data = json.load(load_f)
print('-----', eid)
tree = Tree()
tree.create_node(json_data[0].get("mid"), json_data[0].get("mid"))
for j in range(1, len(json_data)):
try:
tree.create_node(json_data[j].get("mid"),
json_data[j].get("mid"),
parent=json_data[j].get("parent"))
except:
pass
# tree.show()
tree_depth = tree.depth()
if tree_depth > max_depth:
max_depth = tree_depth
#统计各层节点个数
tree_dict = {}
for node in tree.all_nodes_itr():
level = tree.depth(node=node)
if level not in tree_dict:
tree_dict[level] = 1
else:
tree_dict[level] += 1
#统计完
tree_dict_list.append(tree_dict)
tree_levels_count_list = {}
for dict in tree_dict_list:
for i in range(max_depth + 1):
if i in dict:
if i in tree_levels_count_list:
tree_levels_count_list[i] += dict[i]
else:
tree_levels_count_list[i] = dict[i]
print(tree_levels_count_list)
for key in tree_levels_count_list:
print(key, tree_levels_count_list[key] / data_array.shape[0])