def train_multinomial_bayes(request):
prof = request.user.get_profile()
tp = TwitterUserProfile.objects.get(user=prof)
tp.active_classifier = 'MultinomialBayesClassifier'
tp.classifier_version = '0.1'
tp.save()
force_train(tp)
return {'success': 'works?'}
def train_classifier(request):
prof = request.user.get_profile()
tp = TwitterUserProfile.objects.get(user=prof)
force_train(tp)
return {'success': 'works?'}
def cache_timeline_backfill(tp, twitter_tokens, statuses):
""" Backfill cached timeline from the oldest tweet in statuses to
the cached_time in TwitterUserProfile or 72 hours, whichever is sooner"""
api = get_authorized_twython(twitter_tokens)
cutoff_time = datetime.utcnow()-timedelta(hours=72)
if not statuses:
statuses = Status.construct_from_dicts(api.getFriendsTimeline(include_rts=True))
backfill_maxid = min(statuses, key=lambda x: x.id).id
backfill_newestid = max(statuses, key=lambda x: x.id).id
# Maxid and minid indicate contiguous cached status ids
minid = getattr(tp, 'cached_minid', 0)
maxid = getattr(tp, 'cached_maxid', 0)
# No new tweets at all
if backfill_newestid == maxid: return
# Only one page of new tweets - just cache these
if backfill_maxid < maxid:
cache_statuses(statuses, tp)
return
# Cache as far back as 800 tweets or 72 hours worth
num_apicalls = 1
finished = False
total_num_statuses = len(statuses)
while not finished:
print "backfill minid: " + str(maxid)
print "backfill maxid: " + str(backfill_maxid)
recieved_statuses = Status.construct_from_dicts(
api.getFriendsTimeline(count=200, include_rts=True,
max_id=backfill_maxid, min_id=maxid))
num_apicalls += 1
total_num_statuses += len(recieved_statuses)
statuses.extend(recieved_statuses)
oldest_status = min(recieved_statuses, key=lambda x: x.id)
if (oldest_status.created_at < cutoff_time or
oldest_status.id <= maxid or
num_apicalls >= 5): finished = True
else: backfill_maxid = oldest_status.id
# Set new minid, maxid for contiguous cached statuses
if oldest_status.id <= maxid:
tp.cached_minid = minid
else:
tp.cached_minid = oldest_status.id
tp.cached_maxid = backfill_newestid
tp.save()
# If less than 50 rated statuses: force train user classifier
if Rating.objects.filter(user=tp).count() < 50:
force_train(tp)
# print "num apicalls: " + str(num_apicalls)
cache_statuses(statuses, tp)
