def extract_daily_facts(user_1_facts, user_2_facts):
"""
:param user_1_facts:
:param user_2_facts:
:return:
"""
f1 = 0
f2 = 0
f3 = 1.0
f4 = 1.0
num_days = 0
for day in user_1_facts:
if day in user_2_facts:
num_days += 1
facts1 = user_1_facts[day]
facts2 = user_2_facts[day]
jaccard1 = matutils.jaccard(facts1.items(), facts2.items())
jaccard2 = matutils.jaccard(facts1, facts2)
f1 += jaccard1
f2 += jaccard2
f3 = min(f3, jaccard1)
f4 = min(f4, jaccard2)
if num_days > 0:
f1 = float(f1) / num_days
f2 = float(f2) / num_days
else:
f1 = -999
f2 = -999
f3 = -999
f4 = -999
return f1, f2, f3, f4, float(num_days) / (len(user_1_facts) +
len(user_2_facts) - num_days)
vect1 = user_to_topic[u1]
vect2 = user_to_topic[u2]
topic_sim = 1 - spatial.distance.cosine(vect1, vect2)
vec1 = doc2vecmodel.docvecs[u1]
vec2 = doc2vecmodel.docvecs[u2]
doc2vec_sim = 1 - spatial.distance.cosine(vec1, vec2)
bow1 = user_to_words[u1]
bow2 = user_to_words[u2]
lda_bow1 = user_to_lda_bow[u1]
lda_bow2 = user_to_lda_bow[u2]
hellinger_score = matutils.hellinger(
lda_bow1, lda_bow2) # hellinger(lda_bow2, lda_bow1)
cosine_score = matutils.cossim(lda_bow1, lda_bow2)
jaccard_word_score = matutils.jaccard(bow1, bow2)
jaccard_lda_score = matutils.jaccard(lda_bow1, lda_bow2)
outfile.write("{} {} {} {} {} {} {} {} {} {}\n".format(
line.strip(), share_facts,
float(share_facts) / union_facts, tfidf_sim, topic_sim,
hellinger_score, cosine_score, jaccard_word_score,
jaccard_lda_score, doc2vec_sim))
outfile.close()
i += 2
i = 4
while i < len(sys.argv):
outfile = open(sys.argv[i + 1], 'w')
with open(sys.argv[i], 'r') as infile:
for line in tqdm(infile):
splits = line.strip().split()
u1, u2 = splits[0].split(",")
vect1 = tfidf_days[user_to_time_index[u1]].toarray()[0]
vect2 = tfidf_days[user_to_time_index[u2]].toarray()[0]
f1 = 1 - spatial.distance.cosine(vect1, vect2)
f2 = matutils.jaccard(user_to_time[u1], user_to_time[u2])
f3 = matutils.jaccard(user_to_time_id[u1], user_to_time_id[u2])
f4 = compute_euclidean_distance(user_to_hour[u1], user_to_hour[u2])
f5 = compute_euclidean_datetime_distance(user_to_datetime[u1],
user_to_datetime[u2])
f6 = compute_cosine_distance(user_to_hour[u1], user_to_hour[u2])
f7 = compute_cosine_datetime_distance(user_to_datetime[u1],
user_to_datetime[u2])
dist_list, share_time = compute_dist(user_to_hour[u1],
user_to_hour[u2])