def calculate_rate_average_global(raw_data, friend_data):
friend_model = FriendsModel(raw_data,friend_data)
jaccard_sim = CosineSimilarity(friend_model)
roster = friend_model.get_friends_roster()
the_variance = {}
for u_id in roster:
f_ids = friend_model.get_strangers_in_roster(u_id)
f_num = len(friend_model.get_friends(u_id))
sorted_sims = jaccard_sim.get_similarities(u_id,f_ids)[:f_num]
t_ids = []
for u_id, sim in sorted_sims:
t_ids.append(u_id)
mean_rates = friend_model.get_multiple_users_mean_rate(t_ids)
i_rate = friend_model.get_user_rate_mean(u_id)
the_means = []
for ky in mean_rates.keys():
if mean_rates[ky] == 0:
continue
the_means.append(mean_rates[ky])
if not i_rate == 0:
#should always be here
the_means.append(i_rate)
the_variance[u_id] = np.var(the_means)
return the_variance
def get_friends_from_ml100k():
dst_file = "./src/dataset/movie100k/ml100kfriend.dat"
if path.isfile(dst_file):
friend_data = pickle.load(open(dst_file, "rb"))
return friend_data
raw_data = get_moive100k(True)
raw_model = Model(raw_data)
cosine_sim = CosineSimilarity(raw_model)
friend_data = {}
for user_id in raw_model.get_user_ids():
neighbors = cosine_sim.get_similarities(user_id)[:250]
user_ids, x = zip(*neighbors)
user_ids = list(user_ids)
shuffle(user_ids)
# note:
# Randomly choose 150 out of 250 neighbors as friends.
# In such case, systems is able to (possiblly) choose strangers which
# are in top-250 similar users, but with a probability slightly
# smaller than friends selection.
friend_data[user_id] = user_ids[:150]
pickle.dump(friend_data, open(dst_file, "w"), protocol=2)
return friend_data
def calculate_rate_average_global(raw_data, friend_data):
friend_model = FriendsModel(raw_data, friend_data)
jaccard_sim = CosineSimilarity(friend_model)
roster = friend_model.get_friends_roster()
the_variance = {}
for u_id in roster:
f_ids = friend_model.get_strangers_in_roster(u_id)
f_num = len(friend_model.get_friends(u_id))
sorted_sims = jaccard_sim.get_similarities(u_id, f_ids)[:f_num]
t_ids = []
for u_id, sim in sorted_sims:
t_ids.append(u_id)
mean_rates = friend_model.get_multiple_users_mean_rate(t_ids)
i_rate = friend_model.get_user_rate_mean(u_id)
the_means = []
for ky in mean_rates.keys():
if mean_rates[ky] == 0:
continue
the_means.append(mean_rates[ky])
if not i_rate == 0:
#should always be here
the_means.append(i_rate)
the_variance[u_id] = np.var(the_means)
return the_variance
def get_friends_from_ml100k():
dst_file = "./src/dataset/movie100k/ml100kfriend.dat"
if path.isfile(dst_file):
friend_data = pickle.load(open(dst_file, "rb"))
return friend_data
raw_data = get_moive100k(True);
raw_model = Model(raw_data);
cosine_sim = CosineSimilarity(raw_model)
friend_data = {}
for user_id in raw_model.get_user_ids():
neighbors = cosine_sim.get_similarities(user_id)[:250]
user_ids, x = zip(*neighbors)
user_ids = list(user_ids)
shuffle(user_ids)
# note:
# Randomly choose 150 out of 250 neighbors as friends.
# In such case, systems is able to (possiblly) choose strangers which
# are in top-250 similar users, but with a probability slightly
# smaller than friends selection.
friend_data[user_id] = user_ids[:150]
pickle.dump(friend_data,open(dst_file, "w"),protocol=2)
return friend_data
def calculate_cosine_similarity_global(raw_data, friend_data):
friend_model = FriendsModel(raw_data, friend_data)
cosine_sim = CosineSimilarity(friend_model)
c_sims = {}
for u_id in friend_model.get_friends_roster():
f_num = len(friend_model.get_friends(u_id))
target_ids = friend_model.get_strangers_in_roster(u_id)
sorted_sims = cosine_sim.get_similarities(u_id, target_ids)
c_sims[u_id] = sorted_sims[:f_num]
return c_sims
def calculate_cosine_similarity_friends(raw_data, friend_data):
friend_model = FriendsModel(raw_data, friend_data)
cosine_sim = CosineSimilarity(friend_model)
c_sims = {}
for ky in friend_model.get_friends_roster():
f_ids = friend_model.get_friends(ky)
sorted_sims = cosine_sim.get_similarities(ky, f_ids)
c_sims[ky] = sorted_sims
return c_sims
def calculate_cosine_similarity_global(raw_data,friend_data):
friend_model = FriendsModel(raw_data,friend_data)
cosine_sim = CosineSimilarity(friend_model)
c_sims = {}
for u_id in friend_model.get_friends_roster():
f_num = len(friend_model.get_friends(u_id))
target_ids = friend_model.get_strangers_in_roster(u_id)
sorted_sims = cosine_sim.get_similarities(u_id,target_ids)
c_sims[u_id] = sorted_sims[:f_num]
return c_sims
def calculate_cosine_similarity_friends(raw_data,friend_data):
friend_model = FriendsModel(raw_data,friend_data)
cosine_sim = CosineSimilarity(friend_model)
c_sims = {}
for ky in friend_model.get_friends_roster():
f_ids = friend_model.get_friends(ky)
sorted_sims = cosine_sim.get_similarities(ky, f_ids)
c_sims[ky] = sorted_sims
return c_sims
def Generate_Simulating_Data_on_MovieLens(user_id, f_num, t_num):
friend_data = get_friends_from_ml100k()
raw_data = get_moive100k()
friend_model = FriendsModel(raw_data, friend_data)
cosine_sim = CosineSimilarity(friend_model)
fs = Friends_Strangers(cosine_sim, f_num, t_num)
friends = fs.get_rand_friends(user_id)
strangers = fs.get_rand_strangers(user_id)
def generate_simulating_data(raw_data, friend_data, user_id, f_num, t_num,
filelocation):
if not user_id in friend_data:
print "invalid user_id!"
return
if not path.exists(filelocation):
makedirs(filelocation)
friend_model = FriendsModel(raw_data, friend_data)
cosine_sim = CosineSimilarity(friend_model)
fs = Friends_Strangers(cosine_sim, f_num, t_num)
friends = fs.get_rand_friends(user_id)
strangers = fs.get_rand_strangers(user_id)
f_ids, f_sims = zip(*friends)
t_ids, _ = zip(*strangers)
f_sims = np.dot(f_sims, 8).astype(np.int32).tolist() #for HE
#print f_sims
f1 = filelocation + "similarity_" + str(user_id) + "_" + str(
f_num) + ".dat"
save_list_to_file(f1, [f_sims]) #friends similarity
diskfriend = []
for f_id in f_ids:
vec = friend_model.get_dense_user_vector(f_id)
vec = np.dot(vec, 16).astype(np.int32).tolist() #for HE
#vec.tolist()
diskfriend.append(vec)
f2 = filelocation + "friend_" + str(user_id) + "_" + str(f_num) + ".dat"
save_list_to_file(f2, diskfriend) #friends rating data
diskstranger = []
t_ids, __ = zip(*strangers)
for t_id in t_ids:
vec = friend_model.get_dense_user_vector(f_id)
vec = np.dot(vec, 16).astype(np.int32).tolist() #for HE
diskstranger.append(vec)
f3 = filelocation + "stranger_" + str(user_id) + "_" + str(t_num) + ".dat"
save_list_to_file(f3, diskstranger) #friends rating data
def get_similarity_matrix(data_set, friend_data):
i_model = FriendsModel(data_set, friend_data)
similarity = CosineSimilarity(i_model)
return similarity