def cal_tf(uid: str, text: str):
word = tagger.parseToNode(text)
# Parse words
nouns = []
while word:
feature_list = word.feature.split(',')
prototype, vocabulary = feature_list[6], feature_list[0]
if vocabulary == '名詞' and prototype != '*':
nouns.append(prototype)
word = word.next
# Calculate tf
num = len(nouns)
cnt = Counter(nouns)
insert = 'INSERT INTO tf (uid, word, tf) VALUES (%s, %s, %s)'
values = []
for key, val in cnt.items():
tf = val / num
values.append((uid, key, tf))
with GetCursor() as sub_cur:
sub_cur.executemany(insert, values)
def run(self):
with GetCursor() as cur:
query = 'SELECT DISTINCT gameid FROM traindata_yep'
cur.execute(query)
res = list(zip(*cur.fetchall()))[0]
itemList = list(map(str, res))
while True:
if self.lock.acquire():
if list(self.svdNowStep) == [self.avgNowStep.value + 1] * self.processNum:
lSvdItemFeature = [None] * self.processNum
for t in range(self.processNum):
lSvdItemFeature[t] = self.svdItemFeature[t]
lAvgItemFeature = {}
for i in itemList:
lAvgItemFeature.setdefault(i, {})
for j in range(1, self.feature + 1):
value = 0.0
for t in range(self.processNum):
value += lSvdItemFeature[t][i][j]
lAvgItemFeature[i].setdefault(j, value / float(self.processNum))
self.avgItemFeature.update(lAvgItemFeature)
self.avgNowStep.value += 1
print('Process-avg: next step')
if map(lambda x: x < self.stopAt, self.descentRate) == [True] * self.processNum:
self.lock.release()
return
self.lock.release()
time.sleep(1)
def erf(uname: str, k: int):
with GetCursor() as cur:
query = 'SELECT gameid, d1, d2, d3, d4, d5, d6 FROM raw_train_data WHERE userid = \'%s\'' % uname
cur.execute(query)
table = list(map(list, zip(*cur.fetchall())))
lent = len(table[0])
all_ = 0
for line in table[1:]:
all_ += sum(line)
group_time = all_ / k
remind = [0] * lent
aim = []
for line in table[1:]:
new = line
while sum(remind) + sum(new) > group_time * 0.999999:
proportion = (group_time - sum(remind)) / sum(new)
res_list = listadd(remind, listpro(new, proportion))
rate_list = listpro(res_list, 1 / group_time)
aim.append(rate_list)
remind = [0] * lent
new = listpro(new, 1 - proportion)
remind = listadd(remind, new)
rating_list = list(map(list, zip(*aim)))
insert = 'INSERT INTO ts_train_data VALUES (%s, %s, %s)'
for i in range(lent):
ratings = ' '.join(map(lambda x: str(round(x, 3)), rating_list[i]))
cur.execute(insert, (uname, table[0][i], ratings))
def get_tweets(user_id: str):
# https://developer.twitter.com/en/docs/tweets/timelines/api-reference/get-statuses-user_timeline
url = 'https://api.twitter.com/1.1/statuses/user_timeline.json'
params = {'user_id': user_id, 'count': 10}
values = []
for tweet in TwitterConnector().get_json_res(url, params):
values.append((tweet['id_str'], tweet['user']['name'], tweet['text']))
insert = 'INSERT INTO tweets (id, user, text) VALUES (%s, %s, %s)'
with GetCursor() as cur_1:
cur_1.executemany(insert, values)
def get_tweets(screen_name: str, count: int = 100):
url = 'https://api.twitter.com/1.1/statuses/user_timeline.json'
params = {'screen_name': screen_name, 'count': count}
tweets = TwitterConnector().get_json_res(url, params)
uid = tweets[0]['user']['id_str']
tweet_list = []
for tweet in tweets:
tweet_list.append(tweet['text'])
insert = 'INSERT INTO docs (uid, tweets) VALUES (%s, %s)'
with GetCursor() as cur:
cur.execute(insert, (uid, ' '.join(tweet_list)))
def load_data(self):
with GetCursor() as cur:
query = 'SELECT DISTINCT userid FROM traindata_yep'
cur.execute(query)
res = list(zip(*cur.fetchall()))[0]
self.userList = list(map(str, res))
query = 'SELECT DISTINCT gameid FROM traindata_yep'
cur.execute(query)
res = list(zip(*cur.fetchall()))[0]
self.itemList = list(map(str, res))
query = 'SELECT userid, gameid, rating FROM traindata_yep WHERE tg = %d AND rating != 0' % self.tg
cur.execute(query)
for item in cur:
self.train.setdefault(str(item[0]), {})
self.train[str(item[0])][str(item[1])] = float(item[2])
def correct_data(self) -> (list, list):
test_y = []
known = []
with GetCursor() as cur:
for user in self.__user_tpl:
test_list = []
query = 'SELECT gameid FROM date170709 WHERE userid = \'%s\'' % user
cur.execute(query)
for row in cur:
if row[0] in self.__game_dict:
test_list.append(row[0])
test_y.append(test_list)
query = 'SELECT gameid FROM ts_train_data WHERE userid = \'%s\'' % user
cur.execute(query)
known.append(list(zip(*cur.fetchall()))[0])
return test_y, known
def __init__(self, user_limit: int, time_step: int = None):
with GetCursor() as cur:
# Get game list
query = 'SELECT DISTINCT gameid FROM ts_train_data ORDER BY gameid'
cur.execute(query)
self.game_tpl = list(zip(*cur.fetchall()))[0]
self.__game_dict = dict(
(k, v) for v, k in enumerate(self.game_tpl))
self.__game_count = len(self.game_tpl)
# Get user list
query = 'SELECT userid, COUNT(*) AS num FROM raw_train_data GROUP BY userid ORDER BY num DESC LIMIT %d' \
% user_limit
cur.execute(query)
self.__user_tpl = list(zip(*cur.fetchall()))[0]
self.__user_sparse_matrix = None
self.__time_step = time_step
def global_data(self) -> (list, list):
data = []
sign = []
with GetCursor() as cur:
for user in self.__user_tpl:
# Get global rating
query = 'SELECT gameid, rating FROM global_train_data WHERE userid = \'%s\'' % user
cur.execute(query)
rating_list = [0.0] * self.__game_count
sign_list = [0] * self.__game_count
for row in cur:
rating_list[self.__game_dict[row[0]]] = float(row[1])
sign_list[self.__game_dict[row[0]]] = 1
data.append(rating_list)
sign.append(sign_list)
return data, sign
def __do_parse(self):
# Insert values
values = []
with open('../_2_parse_page/%d.txt' % self.__file_num, 'r', encoding='utf-8') as fl:
for line in fl:
# Parse each word
word = tagger.parseToNode(line)
while word:
feature_list = word.feature.split(',')
genkei, hinshi = feature_list[6], feature_list[0]
# Filter out supposed word
if genkei not in ng_genkei and hinshi not in ng_hinshi:
values.append((self.__file_num, genkei, hinshi))
word = word.next
# Execute query
with GetCursor() as cur:
cur.executemany(insert, values)
def __require_usm(self):
if not self.__user_sparse_matrix:
self.__user_sparse_matrix = dict()
with GetCursor() as cur:
for user in self.__user_tpl:
# Get ratings
query = 'SELECT gameid, ratings FROM ts_train_data WHERE userid = \'%s\'' % user
cur.execute(query)
# [row, col, val]
sparse_matrix = [[], [], []]
for rating in cur:
ratings = rating[1].split() # str
self.__ts_count = len(ratings)
for i in range(self.__ts_count):
if float(ratings[i]) > 0:
sparse_matrix[0].append(i)
sparse_matrix[1].append(
self.__game_dict[rating[0]])
sparse_matrix[2].append(float(ratings[i]))
self.__user_sparse_matrix.setdefault(user, sparse_matrix)
from twconnector import TwitterConnector
from dbconnector import GetCursor
# https://developer.twitter.com/en/docs/accounts-and-users/follow-search-get-users/api-reference/get-friends-ids
url = 'https://api.twitter.com/1.1/friends/ids.json'
my_id = '736421314366312448'
values = []
for friend_id in TwitterConnector().get_json_res(url)['ids']:
values.append((my_id, friend_id))
insert = 'INSERT INTO followee (from_id, follow_id) VALUES (%s, %s)'
with GetCursor() as cur:
cur.executemany(insert, values)
def write(games: list, db: str):
with GetCursor() as cur:
data = Counter(games).items()
query = 'INSERT INTO ' + db + ' (gameid, cnt) VALUES (%s, %s)'
cur.executemany(query, data)
