def comps():
global logger_comps
query = json.loads(request.data).get('payload')
logger_comps.info('Comps Query, %s' % query)
cid, tag, start, size = query.get('company'), query.get(
'tag', 0), query.get('start', 0), query.get('size', 5)
if tag == 0:
comps_candidates = dbutil.get_company_comps(g.db, cid)
logger_comps.info(comps_candidates)
results = {
'company': {
'count':
len(comps_candidates),
'data':
map(lambda x: {'id': dbutil.get_company_code(g.db, x)},
comps_candidates)[start:start + size],
'tags':
dbutil.prompt_tag_filter(g.db, comps_candidates)
}
}
else:
tag = dbutil.get_tag_id(g.db, tag)[0]
comps_candidates = dbutil.get_filtered_company_comps(g.db, cid, tag)
results = {
'company': {
'count':
len(comps_candidates),
'data':
map(lambda x: {'id': dbutil.get_company_code(g.db, x)},
comps_candidates)[start:start + size]
}
}
return make_response(jsonify(results))
def match(self, cid):
tags = {
name: weight
for _, name, weight in dbutil.get_company_tags_idname(
self.db, cid, tag_out_type=(11000, 11001, 11002, 11100, 11054))
}
tags = self.transformer.transform(tags)
similarities = cosine_similarity(self.fund_profiles, tags)
similarities = {
self.fund_mapping.get(index): s
for index, s in enumerate([s[0] for s in similarities])
}
# similarities = filter(lambda item: 1020 <= self.fund_rounds.get(item[0]) < 1040, similarities)
# similarities = filter(lambda item: self.activeness.get(item[0]) > 3, similarities)
investor_comps = list(
chain(*[
dbutil.get_company_investors(self.db, comp)
for comp in dbutil.get_company_comps(self.db, cid)
]))
for ic in set(investor_comps):
similarities[ic] = similarities.get(
ic, 0) * investor_comps.count(ic) + 1
similarities = sorted(similarities.items(), key=lambda x: -x[1])
famous = set(dbutil.get_online_investors(self.db))
with codecs.open('dumps/fund', 'w', 'utf-8') as fo:
for iid, weight in similarities:
fo.write('%s\t%s\t%s\t%s\t%s\t%s\n' %
(self.funds.get(iid), weight, iid
in famous, self.fund_rounds.get(iid),
self.activeness.get(iid), self.locations.get(iid)))
def expand(self):
companies = set([
c.companyId
for c in dbutil.get_topic_companies(self.db, self.tpid)
])
if len(companies) < 5:
return
candidates = [
comps.get('candidates', [])
for comps in self.mongo.comps.candidates.find(
{'company': {
'$in': list(companies)
}})
]
candidates = map(lambda x: x[0], chain(*candidates))
candidates = Counter(
[cid for cid in candidates if cid not in companies])
comps = Counter([
cid for cid in chain(
*[dbutil.get_company_comps(self.db, cid) for cid in companies])
if cid not in companies
])
for comp in comps.most_common(min(len(companies) / 2, 50)):
if comp[1] > len(companies) / 5:
dbutil.update_topic_company(self.db,
self.tpid,
comp[0],
confidence=0.51)
for candidate in candidates.most_common(min(len(companies) / 5, 30)):
if candidate[1] > len(companies) * 0.6:
dbutil.update_topic_company(self.db,
self.tpid,
candidate[0],
confidence=0.51)
def company_sample():
db = dbcon.connect_torndb()
mongo = dbcon.connect_mongo()
cids = mongo.task.company.find({
'types': 'visit_local'
}).sort('_id', DESCENDING).limit(500)
cids = sample([r.get('companyId') for r in cids], 100)
with codecs.open('files/sample', 'w', 'utf-8') as fo:
fo.write('\n'.join([str(cid) for cid in cids]))
with codecs.open('files/sample.old', 'w', 'utf-8') as fo:
for cid in cids:
fo.write('%s#%s\n' % (cid, ','.join(
[str(cid2) for cid2 in dbutil.get_company_comps(db, cid)])))