def sampling_cate_topics(regions, size, g_percentages):
""" Sampling poi topics from the database
"""
topics = pd.DataFrame(columns=TOPIC_SCHEMA)
checkins = None
cate_set = set()
for r in regions:
kbase = KnowledgeBase.fromMongo(db.checkin, r['value'])
if checkins is not None:
cate_set = set(kbase.checkins['cid'].unique())
checkins = checkins.append(kbase.checkins, ignore_index=True)
else:
cate_set &= set(kbase.checkins['cid'].unique())
checkins = kbase.checkins
_LOGGER.info('%d checkins loaded for cate_topics', len(checkins))
checkins.drop_duplicates(cols=['pid', 'user'], inplace=True)
for zcate, group in checkins.groupby('z_category'):
cidgroup = [
cid + '\t' + cname
for cid, cname in group[['cid', 'category']].values
if cid in cate_set
]
for gid, g in enumerate(
stratified_samples(cidgroup, g_percentages, size / 9)):
for s in g:
cid, cname = s.split('\t')
for r in regions:
topics = topics.append([{
'topic_id': CATE_ID.next(),
'topic': cname,
'region': r['name'],
'associate_id': cid,
'zcategory': zcate,
'group': gid
}])
_LOGGER.info('# CATE_topics: %d', len(topics))
for zcate, group in checkins.groupby('z_category'):
for r in regions:
topics = topics.append([{
'topic_id': ZCATE_ID.next(),
'topic': zcate,
'region': r['name'],
'associate_id': group['zcid'].values[0],
'zcategory': zcate
}])
_LOGGER.info('# Total Cate_topics: %d', len(topics))
return topics
def sampling_cate_topics(regions, size, g_percentages):
""" Sampling poi topics from the database
"""
topics = pd.DataFrame(columns=TOPIC_SCHEMA)
checkins = None
cate_set = set()
for r in regions:
kbase = KnowledgeBase.fromMongo(db.checkin, r['value'])
if checkins is not None:
cate_set = set(kbase.checkins['cid'].unique())
checkins = checkins.append(kbase.checkins, ignore_index=True)
else:
cate_set &= set(kbase.checkins['cid'].unique())
checkins = kbase.checkins
_LOGGER.info('%d checkins loaded for cate_topics', len(checkins))
checkins.drop_duplicates(cols=['pid', 'user'], inplace=True)
for zcate, group in checkins.groupby('z_category'):
cidgroup = [cid + '\t' + cname
for cid, cname in group[['cid', 'category']].values
if cid in cate_set]
for gid, g in enumerate(stratified_samples(cidgroup,
g_percentages,
size / 9)):
for s in g:
cid, cname = s.split('\t')
for r in regions:
topics = topics.append([{'topic_id': CATE_ID.next(),
'topic': cname,
'region': r['name'],
'associate_id': cid,
'zcategory': zcate,
'group': gid}])
_LOGGER.info('# CATE_topics: %d', len(topics))
for zcate, group in checkins.groupby('z_category'):
for r in regions:
topics = topics.append([{'topic_id': ZCATE_ID.next(),
'topic': zcate,
'region': r['name'],
'associate_id': group['zcid'].values[0],
'zcategory': zcate}])
_LOGGER.info('# Total Cate_topics: %d', len(topics))
return topics
def sampling_poi_topics(region, size, g_percentages):
""" Sampling poi topics from the database
"""
topics = pd.DataFrame(columns=TOPIC_SCHEMA)
kbase = KnowledgeBase.fromMongo(db.checkin, region['value'])
kbase.checkins.drop_duplicates(cols=['pid', 'user'], inplace=True)
for zcate, group in kbase.checkins.groupby('z_category'):
pidgroup = [pid + '\t' + pname
for pid, pname in group[['pid', 'place']].values]
for gid, g in enumerate(stratified_samples(pidgroup,
g_percentages,
size / 9)):
for s in g:
pid, pname = s.split('\t')
topics = topics.append([{'topic_id': POI_ID.next(),
'topic': pname,
'region': region['name'],
'associate_id': pid,
'zcategory': zcate,
'group': gid}])
_LOGGER.info('# POI_topics: %d', len(topics))
return topics
def sampling_poi_topics(region, size, g_percentages):
""" Sampling poi topics from the database
"""
topics = pd.DataFrame(columns=TOPIC_SCHEMA)
kbase = KnowledgeBase.fromMongo(db.checkin, region['value'])
kbase.checkins.drop_duplicates(cols=['pid', 'user'], inplace=True)
for zcate, group in kbase.checkins.groupby('z_category'):
pidgroup = [
pid + '\t' + pname for pid, pname in group[['pid', 'place']].values
]
for gid, g in enumerate(
stratified_samples(pidgroup, g_percentages, size / 9)):
for s in g:
pid, pname = s.split('\t')
topics = topics.append([{
'topic_id': POI_ID.next(),
'topic': pname,
'region': region['name'],
'associate_id': pid,
'zcategory': zcate,
'group': gid
}])
_LOGGER.info('# POI_topics: %d', len(topics))
return topics
