def __init__(self,
use_normalized_data=True,
start_from_zero=False,
learning_rate=0.001):
init_logger(log_file='log/pmf.log', log_level=logging.INFO)
self.exp_id = datetime.today().strftime('%Y%m%d%H%M%S')
self.ratings_file = ratings_file
self.load_data()
self.obs_num = self.ratings_vector.shape[0]
self.use_normalized_data = use_normalized_data
self.start_from_zero = start_from_zero #数据里user_id和item_id是否从0开始，豆瓣是从0开始的
if self.use_normalized_data:
self.generate_normalized_ratings()
self.split_data()
self.learning_rate = learning_rate
self.epsilon = learning_rate
#learning rate
self.lamb = 0.01 #Regularization parameter
self.momentum = 0.8
self.max_epoch = 1000 #iteration
self.feat_num = 10
#uid, vid以observation里出现的uid为准, 如何划分数据也是一个问题
self.user_num = self.ratings_vector[:, 0].max()
self.item_num = self.ratings_vector[:, 1].max()
self.U_shape = (self.user_num, self.feat_num)
self.V_shape = (self.item_num, self.feat_num)
#U: matrix of user features, V: matrix of item features, generated from gaussian distribution
self.U = np.random.standard_normal(self.U_shape)
self.V = np.random.standard_normal(self.V_shape)
def __init__(self, use_normalized_data=True, start_from_zero=False, learning_rate=0.001):
init_logger(log_file='log/pmf.log', log_level=logging.INFO)
self.exp_id = datetime.today().strftime('%Y%m%d%H%M%S')
self.ratings_file = ratings_file
self.load_data()
self.obs_num = self.ratings_vector.shape[0]
self.use_normalized_data = use_normalized_data
self.start_from_zero = start_from_zero#数据里user_id和item_id是否从0开始，豆瓣是从0开始的
if self.use_normalized_data:
self.generate_normalized_ratings()
self.split_data()
self.learning_rate = learning_rate
self.epsilon = learning_rate; #learning rate
self.lamb = 0.01 #Regularization parameter
self.momentum = 0.8
self.max_epoch = 1000 #iteration
self.feat_num = 10
#uid, vid以observation里出现的uid为准, 如何划分数据也是一个问题
self.user_num = self.ratings_vector[:,0].max()
self.item_num = self.ratings_vector[:,1].max()
self.U_shape = (self.user_num, self.feat_num)
self.V_shape = (self.item_num, self.feat_num)
#U: matrix of user features, V: matrix of item features, generated from gaussian distribution
self.U = np.random.standard_normal(self.U_shape)
self.V = np.random.standard_normal(self.V_shape)
def set_logfile(config, args):
log_filename = 'log/fmg_%s_%s_split%s.log' % (
config['dt'], config['exp_type'], config['sn'])
if config['exp_type'] == 'vary_mg':
log_filename = 'log/fmg_%s_%s_split%s_reg%s.log' % (
config['dt'], config['exp_type'], config['sn'], config['reg'])
config['log_filename'] = log_filename
init_logger('', config['log_filename'], logging.INFO, False)
def init_conifg(dt_arg, reg, exp_type, eps, K=10, F=10):
global rating_filename
global logger
global exp_id
global dt
dt = dt_arg
if dt == 'yelp':
rating_filename = 'ratings_filter5'
elif dt in ['yelp-200k', 'yelp-50k', 'yelp-10k', 'yelp-5k', 'yelp-100k']:
rating_filename = 'ratings'
elif dt in ['douban']:
rating_filename = 'ratings'
elif dt == 'cikm-yelp':
rating_filename = 'ratings'
elif dt == 'yelp-sample':
rating_filename = ''
elif dt in ['ml-100k', 'ml-1m', 'ml-10m']:
rating_filename = '%s-rating' % dt
elif dt == 'amazon-app':
rating_filename = 'filter5_ratings_Apps_for_Android'
elif dt in ['amazon-200k', 'amazon-50k', 'amazon-100k', 'amazon-10k', 'amazon-5k']:
rating_filename = 'ratings'
if exp_type == 1:
log_filename = 'log/%s_fm_glasso_once_reg%s_eps%s_K%s_F%s.log' % (dt, reg, eps, K, F)
elif exp_type == 2:
log_filename = 'log/%s_fm_glasso_regv%s_eps%s_K%s_F%s.log' % (dt, reg, eps, K, F)
exp_id = int(time.time())
logger = init_logger('exp_%s' % exp_id, log_filename, logging.INFO, False)
def set_logfile(config, args):
if config.get('reg'):
if args.mg:
motif = re.search('m\d', args.mg).group(0)
log_filename = 'log/%s_%s_%s_%s_reg%s.log' % (
config['dt'], config.get('log_filename'), "glasso", motif,
config.get('reg'))
else:
log_filename = 'log/%s_%s_%s_reg%s.log' % (
config['dt'], config.get('log_filename'), "glasso",
config.get('reg'))
else:
log_filename = 'log/%s_%s_%s_regW%s_regP%s.log' % (
config['dt'], config.get('log_filename'), "glasso",
config.get('reg_W'), config.get('reg_P'))
config['log_filename'] = log_filename
init_logger('exp_%s' % config['exp_id'], config['log_filename'],
logging.INFO, False)
logging.info('******\n%s\n******', config)
import time
import logging
import random
from datetime import datetime
import numpy as np
from sklearn.linear_model import LogisticRegression
from sklearn import preprocessing
from logging_util import init_logger
from rec_dal import RECDAL
from constant import LR_log_file, vali_output_path, test_output_path
dal = RECDAL()
init_logger(log_file=LR_log_file, log_level=logging.INFO, print_console=True)
#sql for offline prediction
train_pos_sql = '''
select l.user_id, l.item_id, s.looks, s.stores, s.carts, s.buys, s.total, s.l3d_looks, s.l3d_stores, s.l3d_carts, l3d_buys, s.l3d_total, s.lc_date_delta,
s.y_looks, s.y_stores, s.y_carts, s.y_buys, s.y_total, s.lc_date_delta, s.item_total, s.item_l3d_total, s.item_yes_total, l.label
from split_20141217_labels as l join split_20141217_stats as s
on l.user_id=s.user_id and l.item_id=s.item_id
where l.label = 1
'''
train_pos_sql = '''
select s.user_id, s.item_id, s.buys, s.l3d_buys, s.y_buys, s.total, s.l3d_total, s.y_total, s.item_total, s.item_l3d_total, s.item_yes_total, s.lc_date_delta, l.label
from split_20141217_labels as l join split_20141217_stats as s
on l.user_id=s.user_id and l.item_id=s.item_id
where l.label = 1
'''
train_neg_sql = '''
import tempfile
import platform
import datetime
import argparse
from configparser import ConfigParser, ExtendedInterpolation
from typing import List, Dict, Tuple
import subprocess
import release_task_reader
from urllib.request import urlretrieve
from urllib.error import HTTPError
from release_task_reader import ReleaseTask
from installer_utils import PackagingError
from runner import exec_cmd, async_exec_cmd
from logging_util import init_logger
log = init_logger(__name__, debug_mode=False)
timestamp = datetime.datetime.fromtimestamp(time.time()).strftime('%Y-%m-%d--%H:%M:%S')
class QtRepositoryLayout:
def __init__(self, root_path: str, license: str, repo_domain: str) -> None:
self.root_path = root_path
self.license = license
self.repo_domain = repo_domain
self.pending = "pending"
self.staging = "staging"
self.production = "production"
# <root_path>/<license>/<pending|staging|production>/<repo_domain>/
# /data/online_repositories/opensource/pending|staging|production/qtsdkrepository/
self.base_repo_path = os.path.join(self.root_path, self.license)
def run_all_epinions():
# run(path_str)
for path_str in ['ratings_only']:
run(path_str)
for path_str in [
'UUB_m1', 'UUB_m2', 'UUB_m3', 'UUB_m4', 'UUB_m5', 'UUB_m6',
'UUB_m7'
]:
for n in range(11):
alpha = n * 0.1
path_str1 = '%s_%s' % (path_str, alpha)
print 'run for ', path_str1
run(path_str1)
if __name__ == '__main__':
global dir_
if len(sys.argv) == 3:
dt = sys.argv[1]
split_num = sys.argv[2]
dir_ = 'data/%s/exp_split/%s/' % (dt, split_num)
log_filename = 'log/%s_mf_feature_geneartion_split%s.log' % (dt,
split_num)
exp_id = int(time.time())
logger = init_logger('exp_%s' % exp_id, log_filename, logging.INFO,
False)
run_all_epinions()
else:
print 'please speficy the data and path_str'
sys.exit(0)
fw = open('data/yelp/samples/grid_res_%s' % filename, 'w+')
fw.write('\n'.join(
['%s\t%s\t%s\t%s\t%s' % (f, k, e, l, r) for f, k, e, l, r in res]))
fw.close()
if __name__ == '__main__':
if len(sys.argv) == 3:
dt = sys.argv[1]
global logger
global dir_
dir_ = 'data/%s/' % dt
exp_id = int(time.time())
log_filename = 'log/%s_mf.log' % dt
logger = init_logger('exp_%s' % str(exp_id), log_filename,
logging.INFO, False)
if int(sys.argv[2]) == 1:
filename = dir_ + 'ratings.txt'
K = 10
eps = 10
lamb = 1
max_iter = 500
exp_rmses = []
for i in range(1):
exp_rmses.append(
run_basedline(filename,
K,
eps,
lamb,
max_iter,
silent_run=False))
