def __init__(self, usr_loc_hist):
# load categories hierarchy
self.ch = hc.load()
# number of users in system
self.n_users = len(usr_loc_hist['user_id'].unique())
# user_venue count matrix
self.uv_cnt = pd.crosstab(usr_loc_hist['user_id'],
usr_loc_hist['business_id'])
# user location history
self.usr_loc_hist = usr_loc_hist.groupby['venue_category_name']
def train(events):
# Load Categories Hierarchy
print 'Loading Categories Hierarchy', datetime.now()
h = ch.load()
categories = h.keys()
mab = ContainerMAB(n_arms=len(categories), n_bins=24)
n_trials = events.shape[0]
returned_cate = [None for i in range(n_trials)]
rewards = [0.0 for i in range(n_trials)]
cumulative_rewards = [0.0 for i in range(n_trials)]
progress = n_trials
for cnt, (r_ind, r) in enumerate(events.iterrows()):
if cnt % (progress/10) == 0:
print ' Progress:', cnt*100/progress, '%', datetime.now()
t = r['datetime']
cate_idx = mab.select_arms(int(t.hour))
category = categories[cate_idx]
returned_cate[cnt] = cate_idx
# Compute Reward
v_category = r['venue_category_id']
if v_category in h:
reward = h.dist_to_LCA(category, v_category, 0)
reward = 1 - (reward/3)**2
else:
reward = 0
mab.update(t.hour, cate_idx, reward)
# Compute Cumulative Reward
rewards[cnt] = reward
if cnt == 0:
cumulative_rewards[cnt] = reward
else:
cumulative_rewards[cnt] = cumulative_rewards[cnt-1] + reward
stats = {
'n_trials': n_trials,
'pulled_arms': returned_cate,
'rewards': rewards,
'cumulative_rewards': cumulative_rewards
}
return mab, stats
def train(events):
# Load Categories Hierarchy
print 'Loading Categories Hierarchy', datetime.now()
h = ch.load()
categories = h.keys()
mab = ContainerMAB(n_arms=len(categories), n_bins=24)
n_trials = events.shape[0]
returned_cate = [None for i in range(n_trials)]
rewards = [0.0 for i in range(n_trials)]
cumulative_rewards = [0.0 for i in range(n_trials)]
progress = n_trials
for cnt, (r_ind, r) in enumerate(events.iterrows()):
if cnt % (progress / 10) == 0:
print ' Progress:', cnt * 100 / progress, '%', datetime.now()
t = r['datetime']
cate_idx = mab.select_arms(int(t.hour))
category = categories[cate_idx]
returned_cate[cnt] = cate_idx
# Compute Reward
v_category = r['venue_category_id']
if v_category in h:
reward = h.dist_to_LCA(category, v_category, 0)
reward = 1 - (reward / 3)**2
else:
reward = 0
mab.update(t.hour, cate_idx, reward)
# Compute Cumulative Reward
rewards[cnt] = reward
if cnt == 0:
cumulative_rewards[cnt] = reward
else:
cumulative_rewards[cnt] = cumulative_rewards[cnt - 1] + reward
stats = {
'n_trials': n_trials,
'pulled_arms': returned_cate,
'rewards': rewards,
'cumulative_rewards': cumulative_rewards
}
return mab, stats
def __init__(self, usr_loc_hist):
# Load categories hierarchy
self.ch = hc.load()
# number of users in system
self.n_users = len(usr_loc_hist['user_id'].unique())
# user-venue count matrix
self.uv_cnt = pd.crosstab(usr_loc_hist['user_id'],
usr_loc_hist['venue_id'])
# user location history
self.usr_loc_hist = usr_loc_hist.groupby(['venue_category_id'])
# venues in categories
self.venues = {}
# user's preference weight
self.ucw = {}
# category experts
self.experts = {}
# Train
self.offline()
def __init__(self, usr_loc_hist):
# Load categories hierarchy
self.ch = hc.load()
# number of users in system
self.n_users = len(usr_loc_hist['user_id'].unique())
# user-venue count matrix
self.uv_cnt = pd.crosstab(usr_loc_hist['user_id'],
usr_loc_hist['venue_id'])
# user location history
self.usr_loc_hist = usr_loc_hist.groupby(['venue_category_id'])
# venues in categories
self.venues = {}
# user's preference weight
self.ucw = {}
# category experts
self.experts = {}
# Train
self.offline()
import pickle
import pandas as pd
import categories_hierarchy as hc
import json
# Phoenix_rating_hist,LV_rating_hist = pickle.load(open('user_rating_hist_after_filter.p','rb'))
hierarchy = hc.load()
file = open('/Users/xiwang/git-code/Dataset/poi_data/yelp_dataset_round11/business.json', 'r')
category_alias_map = {}
with open('./data/categories.json', 'rb') as f:
categories = json.load(f)
for category in categories:
category_alias_map[category['title']] = category['alias']
f.close()
location_category = {}
unmatched_cat = set()
count_business = 0
count_no_cat_business = 0
for line in file.readlines():
location = json.loads(line)
min_level = 10
count_business += 1
if len(location['categories']) > 0:
general_cat = ' '
for category in location['categories']:
if category in category_alias_map and hierarchy.get_level_num(category_alias_map[category]) < min_level:
general_cat = category_alias_map[category]
else:
# print('No Match Category Exception: ', category)
