def portrayal_thread(user, task_id):
tweet_list = tweets_crawler.get_user_all_timeline_return(
screen_name=user['screen_name'])
user['tweets'] = tweet_list
out = user_profile.user_profile(user)
db = MongoDB().connect()
collect = db['typical']
cur_user = collect.find_one({'_id': long(out['user_id'])})
out['_id'] = long(out['user_id'])
del out['user_id']
if cur_user == None:
collect.insert_one(out)
th = threading.Thread(target=relation_thread, args=(out, ))
th.start()
else:
collect.delete_one({'_id': long(out['_id'])})
collect.insert_one(out)
with app.app_context():
TypicalCharacter.query.filter(TypicalCharacter.id == task_id).update({
'finished_at':
time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time()))
})
def classify_career():
db = MongoDB().connect()
users = db['typical'].find()
n = 0
err_dict = {}
for user in users:
tweets = user['tweets']
text = user['description']
for tweet in tweets:
text += ' ' + tweet['text']
text = preprocess.preprocess(text)
res = classify.exe_classify(text)
if err_dict.has_key(res[0]):
err_dict[res[0]] += 1
else:
err_dict[res[0]] = 1
if res[0] == user['category']:
n += 1
print err_dict
print n
def task_detail(task_id):
task = Task.query.filter(Task.id == task_id).first()
user = Admin.query.filter(Admin.userid == task.userid).first()
res = {
'id': task.id,
'task_name': task.task_name,
'user_name': user.username,
'created_at': task.created_at,
'finished_at': task.finished_at,
'screen_name': task.search_name,
'remark': task.remark,
'thread_num': task.thread_num,
'deepth': task.deepth,
'extension': str(task.extension),
'tweet_num': task.tweet_num,
'basicinfo_num': task.basicinfo_num,
'type': 'File Upload' if task.is_file else 'Breadth First Extension',
'search_type': task.search_type,
'file_content': task.file_content,
'basicinfo_num_finished': '',
'tweet_num_finished': ''
}
if '4' in task.search_type:
sql = "select count(*) from task_%s" % task.id
bsc = db.session.execute(sql).first()
res['basicinfo_num_finished'] = bsc[0]
if '1' in task.search_type or task.search_type == '':
md = MongoDB().connect()
collect = md["task_%s" % task.id]
res['tweet_num_finished'] = collect.find().count()
return render_template('task_detail.html', task=res)
def modify_category():
user_id = request.form['user_id']
category = request.form['category']
categories = [
"Entertainment", "Agriculture", "Sports", "Religion", "Military",
"Politics", "Education", "Technology", "Economy"
]
if not user_id or category not in categories:
return jsonify({'status': 0})
db = MongoDB().connect()
collect = db['typical']
collect.update_one({'_id': long(user_id)},
{"$set": {
"category": category
}})
node = selector.select("Typical", user_id=long(user_id)).first()
node['category'] = category
graph.push(node)
return jsonify({'status': 1})
def test_insert():
db = MongoDB('test')
doc = 'test'
data = {'test': 123}
r = db.insert(doc, data=data)
assert r
assert isinstance(r, ObjectId)
c = db.query(doc) \
.find_one({'_id': r})
assert c['test'] == 123
def calc_sentiment_score():
sentiment_dict = dict(map(lambda (k,v): (k,int(v)),
[ line.split('\t') for line in open("portrayal/sentiment_classify/data/sentiment_words1.txt") ]))
db = MongoDB().connect()
users = db['typical'].find({'screen_name': 'EP_Agriculture'}).limit(1)
for user in users:
final_sentiment, psy_with_time1, psy_with_time2, psy_with_count1, psy_with_count2 = sentiment_classify.exe_sentiment_classify(user['tweets'])
# print tags
db['typical'].update({'_id': user['_id']}, {"$set": {"psy": final_sentiment, "psy_with_time1": psy_with_time1, "psy_with_time2": psy_with_time2, "psy_with_count1": psy_with_count1, "psy_with_count2": psy_with_count2}})
def test_update():
db = MongoDB('test')
doc = 'test'
data = {'test': 123}
db.insert(doc, data=data)
c = db.query(doc) \
.find_one({'test': 123})
r = db.update(doc, c=c, data={'test': 999})
assert r['ok'] == True
c2 = db.query(doc) \
.find_one({'_id': c['_id']})
assert c2['test'] == 999
def calc_sentiment():
db = MongoDB().connect()
users = db['typical'].find()
for user in users:
try:
final_sentiment, psy_with_time1, psy_with_time2, psy_with_count1, psy_with_count2 = sentiment_classify.exe_sentiment_classify(user['tweets'])
except Exception as e:
print user['_id']
print e
continue
db['typical'].update({'_id': user['_id']}, {"$set": {"psy": final_sentiment, "psy_with_time1": psy_with_time1, "psy_with_time2": psy_with_time2, "psy_with_count1": psy_with_count1, "psy_with_count2": psy_with_count2}})
def download_user_xml(user_id):
db = MongoDB().connect()
collect = db['typical']
user = collect.find_one({'_id': long(user_id)})
file_name = generate_user_xml(user)
response = make_response(send_file(file_name))
response.headers[
"Content-Disposition"] = "attachment; filename=%s.xml" % user[
'screen_name']
return response
def update_attr():
graph = Neo4j().connect()
mongo = MongoDB().connect()
tus = mongo['typical'].find({}, {
'name': 1,
'category': 1,
'followers_count': 1,
'location': 1,
'utc_offset': 1,
'statuses_count': 1,
'description': 1,
'friends_count': 1,
'psy': 1,
'verified': 1,
'lang': 1,
'favourites_count': 1,
'screen_name': 1,
'influence_score': 1,
'created_at': 1,
'time_zone': 1,
'protected': 1,
'activity': 1
})
for item in tus:
node = graph.find_one("Typical",
property_key="user_id",
property_value=item['_id'])
node['name'] = item['name']
node['category'] = item['category']
node['followers_count'] = item['followers_count']
node['location'] = item['location']
node['utc_offset'] = item['utc_offset']
node['statuses_count'] = item['statuses_count']
node['description'] = item['description']
node['friends_count'] = item['friends_count']
node['psy'] = item['psy']
node['verified'] = item['verified']
node['lang'] = item['lang']
node['favourites_count'] = item['favourites_count']
node['screen_name'] = item['screen_name']
node['influence_score'] = item['influence_score']
node['created_at'] = item['created_at']
node['time_zone'] = item['time_zone']
node['protected'] = item['protected']
node['activity'] = item['activity']
graph.push(node)
def extract_interset():
db = MongoDB().connect()
users = db['typical'].find()
for u in users:
text = ''
for item in u['tweets']:
text += item['text'] + ' '
try:
tags = interest_extract.extract_tags(text, u['description'])
except Exception as e:
print u['_id']
print e
continue
db['typical'].update({'_id': u['_id']}, {"$set": {"interest_tags": tags}})
def download_interest_tags(user_id):
db = MongoDB().connect()
collect = db['typical']
user = collect.find_one({'_id': long(user_id)}, {
'interest_tags': 1,
'screen_name': 1
})
file_name = tag_cloud.generate_tag_cloud(user['interest_tags'],
user['_id'])
response = make_response(send_file(file_name))
response.headers[
"Content-Disposition"] = "attachment; filename=%s_interest_tags.png" % user[
'screen_name']
return response
def typical_character_newdelete():
user_id = request.form['user_id']
if not user_id:
return jsonify({'status': 0})
TypicalCharacter.query.filter(TypicalCharacter.user_id == user_id).delete()
db = MongoDB().connect()
collect = db['typical']
collect.delete_one({'_id': long(user_id)})
collect = db['relation']
collect.delete_one({'_id': long(user_id)})
cql = '''MATCH (a:Typical{user_id:%d})-[r:following]-(b) DETACH delete a, r''' % long(
user_id)
graph.run(cql)
return jsonify({'status': 1})
def typical_data_statistics():
db = MongoDB().connect()
collect = db['typical']
users = collect.find({}, {'category': 1, '_id': 0, 'influence_score': 1})
category = {
'Politics': 0,
'Religion': 0,
'Military': 0,
'Economy': 0,
'Technology': 0,
'Education': 0,
'Agriculture': 0,
'Entertainment': 0,
'Sports': 0
}
influence = {}
for i in range(15):
influence[str(i * 10)] = 0
for item in users:
category[item['category']] += 1
s = str(int(item['influence_score'] / 10) * 10)
if int(s) >= 150:
if not influence.has_key(s):
influence[s] = 1
else:
influence[s] += 1
else:
influence[s] += 1
return render_template('portrayal/typical_data_statistics.html',
category=category,
influence=influence)
def create_relation():
graph = Neo4j().connect()
mongo = MongoDB().connect()
# 清空neo4j数据库
# graph = graph.delete_all()
tus = mongo['typical'].find({}, {'_id': 1})
for item in tus:
# 创建用户节点
user = Node("Typical", user_id=item['_id'])
graph.create(user)
tus = mongo['relation'].find({}, {'_id': 1})
user_list = map(lambda item: item['_id'], tus)
# 创建用户节点之间的关系
for user_id in user_list:
friends = mongo['relation'].find_one({'_id': user_id})
friends = set(friends['friends'])
node1 = graph.find_one("Typical",
property_key="user_id",
property_value=user_id)
for user_id1 in user_list:
if user_id1 == user_id:
continue
if user_id1 in friends:
node2 = graph.find_one("Typical",
property_key="user_id",
property_value=user_id1)
following = Relationship(node1, 'following', node2)
graph.create(following)
def test_query():
db = MongoDB('test')
doc = 'test'
r = db.query(doc)
assert r
assert r == db.collection[doc]
def update_user_category():
db = MongoDB().connect()
users = db['typical'].find({}, {'_id': 1, 'screen_name': 1, 'category': 1, 'category_score': 1})
count = 0
category_name = ['Politics', 'Religion', 'Military', 'Economy', 'Technology', 'Education', 'Agriculture', 'Entertainment', 'Sports']
users_temp = []
for item in users:
sorted_list = sorted(item['category_score'].iteritems(), key = lambda asd:asd[1], reverse = True)
if sorted_list[0][1] > 2 * sorted_list[1][1] or sorted_list[0][1] - sorted_list[1][1] > 50:
if sorted_list[0][0] != item['category']:
count += 1
continue
score_differ = (2 * sorted_list[0][1] - sorted_list[1][1] - sorted_list[-1][1]) / 2
relation_dict = {
sorted_list[0][0]: 0,
sorted_list[1][0]: 0,
sorted_list[2][0]: 0,
sorted_list[3][0]: 0
}
# for name in category_name:
# relation_dict[name] = 0
cql = '''MATCH(a{user_id:%s})-[:following]->(f) return distinct f.user_id as user_id''' % (item['_id'])
res = graph.data(cql)
for f in res:
user = db['typical'].find_one({'_id': f['user_id']}, {'category_score': 1})
category_score = user['category_score']
max_category = max(category_score, key = lambda x: category_score[x])
if max_category in relation_dict:
relation_dict[max_category] += 1
cql = '''MATCH(a{user_id:%s})<-[:following]-(f) return distinct f.user_id as user_id''' % (item['_id'])
res = graph.data(cql)
for f in res:
user = db['typical'].find_one({'_id': f['user_id']}, {'category_score': 1})
category_score = user['category_score']
max_category = max(category_score, key = lambda x: category_score[x])
if max_category in relation_dict:
relation_dict[max_category] += 1
relation_total = 0
for ri in relation_dict:
relation_total += relation_dict[ri]
if relation_total < 10:
if sorted_list[0][0] != item['category']:
count += 1
continue
for ri in relation_dict:
item['category_score'][ri] += round(score_differ * relation_dict[ri] / relation_total, 2)
users_temp.append({'_id':item['_id'], "category_score": item['category_score']})
s1 = sorted_list[0][0]
sorted_list = sorted(item['category_score'].iteritems(), key = lambda asd:asd[1], reverse = True)
# if sorted_list[0][0] == item['category'] and s1 != item['category']:
# print item['screen_name']
if sorted_list[0][0] != item['category']:
count += 1
print count
for item in users_temp:
db['typical'].update({'_id': item['_id']}, {"$set": {"category_score": item['category_score']}})
# print user['screen_name']
# # break
# print count
# for tt in user['tweets']:
# print tt['text']
# extract_interset()
# calc_sentiment()
# calc_sentiment_score()
# sentiment_dict_test()
# try:
# words = word_tokenize("What a beautiful sunday . happy")
# print nltk.pos_tag(words)
# except Exception as e:
# print e
db = MongoDB().connect()
users = db['typical'].find()
count = 1
data_set = {
'retweet_favorite_rate': [],
'fans_retweet_rate': [],
'fans_favorite_rate': []
}
for user in users:
tweets = user['tweets']
fans = user['followers_count']
# if fans > 2000000:
# continue
count += 1
def typical_category_statistics():
db = MongoDB().connect()
collect = db['typical']
total_count = collect.find({}, {}).count()
users = collect.find({}, {'category': 1, '_id': 0, 'category_score': 1})
category_name = [
'Politics', 'Religion', 'Military', 'Economy', 'Technology',
'Education', 'Agriculture', 'Entertainment', 'Sports'
]
category = {}
for item in category_name:
category[item] = {
'count': 0,
'error_count': 0,
'sub_error_count': 0,
'error_classified_count': 0,
'error_distribution': {}
}
for name in category_name:
category[item]['error_distribution'][name] = 0
for item in category:
del category[item]['error_distribution'][item]
error_count = 0
sub_error_count = 0
for item in users:
category[item['category']]['count'] += 1
category_score = item['category_score']
max_category = max(category_score, key=category_score.get)
if max_category != item['category']:
category[item['category']]['error_distribution'][max_category] += 1
category[max_category]['error_classified_count'] += 1
error_count += 1
category[item['category']]['error_count'] += 1
category_score[max_category] = 0
max_category = max(category_score, key=category_score.get)
if max_category != item['category']:
sub_error_count += 1
category[item['category']]['sub_error_count'] += 1
for item in category:
correct_count = category[item]['count'] - category[item]['error_count']
category[item]['correct_count'] = correct_count
category[item]['recall'] = round(
correct_count * 1.0 / category[item]['count'], 4)
category[item]['precision'] = round(
correct_count * 1.0 /
(correct_count + category[item]['error_classified_count']), 4)
category[item]['f_score'] = round(
2 * category[item]['recall'] * category[item]['precision'] /
(category[item]['recall'] + category[item]['precision']), 4)
category[item]['accuracy'] = round(
(correct_count + total_count - category[item]['count'] -
category[item]['error_classified_count']) * 1.0 / total_count, 4)
category[item]['easy_wrong_category'] = max(
category[item]['error_distribution'],
key=category[item]['error_distribution'].get)
average_recall = 0
average_precision = 0
average_fscore = 0
average_accuracy = 0
for item in category:
average_recall += category[item]['recall'] * category[item][
'count'] / total_count
average_precision += category[item]['precision'] * category[item][
'count'] / total_count
average_fscore += category[item]['f_score'] * category[item][
'count'] / total_count
average_accuracy += category[item]['accuracy'] * category[item][
'count'] / total_count
return render_template('portrayal/typical_category_statistics.html', category = category, total_count = total_count, error_count = error_count, \
average_accuracy = round(average_accuracy, 4), sub_error_count = sub_error_count, average_recall = round(average_recall, 4), average_precision = round(average_precision, 4), average_fscore = round(average_fscore, 4))
def relation_thread(user):
user_id = user['_id']
db = MongoDB().connect()
collect = db['relation']
user_id = long(user_id)
typical_user = collect.find_one({'_id': user_id})
if typical_user:
return
user_node = Node("Typical",
user_id=user_id,
name=user['name'],
category=user['category'],
followers_count=user['followers_count'],
location=user['location'],
utc_offset=user['utc_offset'],
statuses_count=user['statuses_count'],
description=user['description'],
friends_count=user['friends_count'],
psy=user['psy'],
verified=user['verified'],
lang=user['lang'],
favourites_count=user['favourites_count'],
screen_name=user['screen_name'],
influence_score=user['influence_score'],
created_at=user['created_at'],
time_zone=user['time_zone'],
protected=user['protected'],
activity=user['activity'])
graph.create(user_node)
cursor = -1
friends = []
while cursor != 0:
out = relation_crawler.get_friendids_paged_sleep(user_id=user_id,
cursor=cursor,
count=5000)
if not out:
break
friends = friends + out[2]
cursor = out[0]
collect.insert_one({'_id': user_id, 'friends': friends})
tus = collect.find()
friends = set(friends)
for item in tus:
if item['_id'] == user_id:
continue
friend_node = selector.select("Typical", user_id=item['_id']).first()
# friend_node1 = graph.find_one("Typical",
# property_key = "user_id",
# property_value = item['_id'])
if item['_id'] in friends:
following = Relationship(user_node, 'following', friend_node)
graph.create(following)
if user_id in set(item['friends']):
following = Relationship(friend_node, 'following', user_node)
graph.create(following)
sorted_list = sorted(user['category_score'].iteritems(),
key=lambda asd: asd[1],
reverse=True)
if sorted_list[0][1] > 2 * sorted_list[1][1] or sorted_list[0][
1] - sorted_list[1][1] > 50:
return
score_differ = (2 * sorted_list[0][1] - sorted_list[1][1] -
sorted_list[-1][1]) / 2
category_name = [
'Politics', 'Religion', 'Military', 'Economy', 'Technology',
'Education', 'Agriculture', 'Entertainment', 'Sports'
]
relation_dict = {}
for name in category_name:
relation_dict[name] = 0
cql = '''MATCH(a{user_id:%s})-[:following]->(f) return distinct f.user_id as user_id, f.category as category''' % (
user['_id'])
res = graph.data(cql)
for f in res:
relation_dict[f['category']] += 1
cql = '''MATCH(a{user_id:%s})<-[:following]-(f) return distinct f.user_id as user_id, f.category as category''' % (
user['_id'])
res = graph.data(cql)
for f in res:
relation_dict[f['category']] += 1
relation_total = 0
for ri in relation_dict:
relation_total += relation_dict[ri]
if relation_total < 10:
return
for ri in relation_dict:
user['category_score'][ri] += round(
score_differ * relation_dict[ri] / relation_total, 2)
category_score = user['category_score']
max_category = max(category_score, key=lambda x: category_score[x])
db['typical'].update(
{'_id': user['_id']},
{"$set": {
"category_score": category_score,
"category": max_category
}})
node = selector.select("Typical", user_id=long(user['_id'])).first()
node['category'] = max_category
graph.push(node)
def typical_character_list_detail():
data = json.loads(request.form['aoData'])
for item in data:
if item['name'] == 'sSearch':
s_search = item['value'].strip()
if item['name'] == 'iDisplayLength':
data_length = item['value']
if item['name'] == 'iDisplayStart':
data_start = item['value']
data_length = int(data_length)
db = MongoDB().connect()
collect = db['typical']
field = {
'screen_name': 1,
'name': 1,
'friends_count': 1,
'followers_count': 1,
'statuses_count': 1,
'influence_score': 1,
'category': 1
}
if s_search == '':
users = collect.find({}, field).skip(data_start).limit(data_length)
count = collect.find({}, {'_id': 1}).count()
else:
pattern = re.compile(".*" + s_search + ".*")
query = {
"$or": [{
"name": pattern
}, {
"category": pattern
}, {
"interest_tags": pattern
}, {
"screen_name": pattern
}, {
"description": pattern
}, {
"location": pattern
}, {
"name": pattern
}]
}
users = collect.find(query, field).skip(data_start).limit(data_length)
count = collect.find(query, {'_id': 1}).count()
res = []
for u in users:
res.append({
"_id": str(u['_id']),
"screen_name": u['screen_name'],
"name": u['name'],
"friends_count": u['friends_count'],
"followers_count": u['followers_count'],
"statuses_count": u['statuses_count'],
"category": u['category'],
"influence_score": round(u['influence_score'], 3)
})
return jsonify({'aaData': res, 'iTotalDisplayRecords': count})
def typical_character_detail(user_id):
mdb = MongoDB().connect()
collect = mdb['typical']
user = collect.find_one({'_id': long(user_id)}, {'tweets': 0})
user['ratio'] = user[
'followers_count'] if not user['friends_count'] else round(
user['followers_count'] * 1.0 / user['friends_count'], 2)
user['created_at'] = time.strftime(
'%Y-%m-%d %H:%M:%S',
time.strptime(user['created_at'].replace('+0000 ', '')))
user['crawler_date'] = str(user['crawler_date']).split(" ")[0]
user['interest_tags'] = user['interest_tags'].replace(',', ', ')
user['interest_tags'] = re.sub(
r'#(\w+)',
"<a href='https://www.twitter.com/hashtag/\g<1>' target='_blank'>#\g<1></a>",
user['interest_tags'])
s = ''
for item in user['activity_list']:
s += "," + str(item)
user['activity_list'] = s[1:]
s = ''
for item in user['psy_with_time1']:
s += "," + str(item)
user['psy_with_time1'] = s[1:]
s = ''
for item in user['psy_with_time2']:
s += "," + str(item)
user['psy_with_time2'] = s[1:]
s = ''
for item in user['psy_with_count1']:
s += "," + str(item)
user['psy_with_count1'] = s[1:]
s = ''
for item in user['psy_with_count2']:
s += "," + str(item)
user['psy_with_count2'] = s[1:]
s = ''
s1 = ''
for item in user['category_score']:
s += "," + item
s1 += "," + str(user['category_score'][item])
user['category_score_keys'] = s[1:]
user['category_score_values'] = s1[1:]
user_tweets = collect.aggregate([{"$match": {'_id': long(user_id)}}, \
{"$project": {"length": {"$size": "$tweets"}, "first": {"$slice": ["$tweets.created_at", 0, 1]},\
"last": {"$slice": ["$tweets.created_at", -1]}}}])
for item in user_tweets:
user['tweets_count'] = item['length']
user['tweets_start_time'] = time.strftime(
'%Y-%m-%d %H:%M:%S',
time.strptime(item['first'][0].replace('+0000 ', '')))
user['tweets_end_time'] = time.strftime(
'%Y-%m-%d %H:%M:%S',
time.strptime(item['last'][0].replace('+0000 ', '')))
break
get_image(user['profile_image_url'], user['screen_name'])
related_users = collect.find(
{
'category': user['category'],
'_id': {
"$ne": user['_id']
}
}, {
'screen_name': 1,
'name': 1,
'profile_image_url': 1
}).limit(10)
ru_arr = []
for ru in related_users:
ru_arr.append({
'_id': str(ru['_id']),
'screen_name': ru['screen_name'],
'name': ru['name']
})
get_image(ru['profile_image_url'], ru['screen_name'])
return render_template('portrayal/typical_character_detail.html',
user=user,
related_users=ru_arr)