def count_items_in_common(self):
"""
Counts the number of items each user has in common with every other user
and stores the results in a dictionary.
:return: a dictionary with the count of the number of times users have
a certain number of items in common. For example, the dictionary
{0:4, 1:10, 2:6, 3:3, 4:1} means that there a 4 users who have 0 items
in common with the rest of users, there are 10 users who have rated 1
item in common with the rest of users, 6 users who have rated 6 items in
common with the rest of the users, and so on.
"""
common_item_counts = {}
user_dictionary = extractor.initialize_users(self.reviews, False)
for i in range(self.num_users):
for j in range(i + 1, self.num_users):
user1 = self.user_ids[i]
user2 = self.user_ids[j]
num_common_items = len(
extractor.get_common_items(user_dictionary, user1, user2))
if num_common_items in common_item_counts:
common_item_counts[num_common_items] += 1
else:
common_item_counts[num_common_items] = 1
return common_item_counts
def count_items_in_common(self):
"""
Counts the number of items each user has in common with every other user
and stores the results in a dictionary.
:return: a dictionary with the count of the number of times users have
a certain number of items in common. For example, the dictionary
{0:4, 1:10, 2:6, 3:3, 4:1} means that there a 4 users who have 0 items
in common with the rest of users, there are 10 users who have rated 1
item in common with the rest of users, 6 users who have rated 6 items in
common with the rest of the users, and so on.
"""
common_item_counts = {}
user_dictionary = extractor.initialize_users(self.reviews, False)
for i in range(self.num_users):
for j in range(i + 1, self.num_users):
user1 = self.user_ids[i]
user2 = self.user_ids[j]
num_common_items = len(extractor.get_common_items(user_dictionary, user1, user2))
if num_common_items in common_item_counts:
common_item_counts[num_common_items] += 1
else:
common_item_counts[num_common_items] = 1
return common_item_counts
def calculate_users_similarity(self, user_dictionary, user1, user2):
common_items = extractor.get_common_items(user_dictionary, user1,
user2)
if not common_items:
return None
if self._min_common_items is not None and len(
common_items) < self._min_common_items:
return None
user1_overall_ratings = user_dictionary[user1].item_ratings
user1_multi_ratings = user_dictionary[user1].item_multi_ratings
user2_overall_ratings = user_dictionary[user1].item_ratings
user2_multi_ratings = user_dictionary[user2].item_multi_ratings
similarity_sum = 0.
for item in common_items:
user1_item_ratings = list(user1_multi_ratings[item])
user1_item_ratings.insert(0, user1_overall_ratings[item])
user2_item_ratings = list(user2_multi_ratings[item])
user2_item_ratings.insert(0, user2_overall_ratings[item])
similarity_sum += similarity_calculator.calculate_similarity(
user1_item_ratings, user2_item_ratings,
self._similarity_metric)
similarity = similarity_sum / len(common_items)
return similarity
def calculate_pearson_similarity(self, user1, user2):
common_items =\
extractor.get_common_items(self.user_dictionary, user1, user2)
if not common_items:
return None
numerator = 0
denominator1 = 0
denominator2 = 0
user1_average = self.user_dictionary[user1].average_overall_rating
user2_average = self.user_dictionary[user2].average_overall_rating
for item in common_items:
user1_rating = self.user_dictionary[user1].item_ratings[item]
user2_rating = self.user_dictionary[user2].item_ratings[item]
# print('user average', user1_average)
numerator +=\
(user1_rating - user1_average) * (user2_rating - user2_average)
denominator1 += (user1_rating - user1_average) ** 2
denominator2 += (user2_rating - user2_average) ** 2
denominator = math.sqrt(denominator1 * denominator2)
if denominator == 0:
return 0
return numerator / denominator
def calculate_cosine_similarity(self, user1, user2):
common_items =\
extractor.get_common_items(self.user_dictionary, user1, user2)
if not common_items:
return None
numerator = 0
denominator1 = 0
denominator2 = 0
for item in common_items:
user1_rating = self.user_dictionary[user1].item_ratings[item]
user2_rating = self.user_dictionary[user2].item_ratings[item]
numerator += user1_rating * user2_rating
denominator1 += user1_rating ** 2
denominator2 += user2_rating ** 2
denominator = math.sqrt(denominator1) * math.sqrt(denominator2)
# if denominator == 0:
# pass
return numerator / denominator
def calculate_users_similarity(self, user_dictionary, user1, user2):
common_items = extractor.get_common_items(user_dictionary, user1, user2)
if not common_items:
return None
if self._min_common_items is not None and len(
common_items) < self._min_common_items:
return None
user1_overall_ratings = user_dictionary[user1].item_ratings
user1_multi_ratings = user_dictionary[user1].item_multi_ratings
user2_overall_ratings = user_dictionary[user1].item_ratings
user2_multi_ratings = user_dictionary[user2].item_multi_ratings
similarity_sum = 0.
for item in common_items:
user1_item_ratings = list(user1_multi_ratings[item])
user1_item_ratings.insert(0, user1_overall_ratings[item])
user2_item_ratings = list(user2_multi_ratings[item])
user2_item_ratings.insert(0, user2_overall_ratings[item])
similarity_sum += similarity_calculator.calculate_similarity(
user1_item_ratings, user2_item_ratings, self._similarity_metric)
similarity = similarity_sum / len(common_items)
return similarity
def calculate_user_similarity(self, user1, user2, threshold):
common_items = extractor.get_common_items(
self.user_dictionary, user1, user2)
if not common_items:
return None
filtered_items = {}
for item in common_items:
context1 = self.user_dictionary[user1].item_contexts[item]
context2 = self.user_dictionary[user2].item_contexts[item]
context_similarity = context_utils.get_context_similarity(
context1, context2, self.context_rich_topics)
if context_similarity > threshold:
filtered_items[item] = context_similarity
numerator = 0
denominator1 = 0
denominator2 = 0
denominator3 = 0
user1_average = self.user_dictionary[user1].average_overall_rating
user2_average = self.user_dictionary[user2].average_overall_rating
for item in filtered_items.keys():
context_similarity = filtered_items[item]
user1_rating = self.user_dictionary[user1].item_ratings[item]
user2_rating = self.user_dictionary[user2].item_ratings[item]
numerator +=\
(user1_rating - user1_average) *\
(user2_rating - user2_average) *\
context_similarity
denominator1 += (user1_rating - user1_average) ** 2
denominator2 += (user2_rating - user2_average) ** 2
denominator3 += context_similarity ** 2
denominator = math.sqrt(denominator1 * denominator2 * denominator3)
if denominator == 0:
return None
return numerator / denominator
def calculate_user_similarity(self, user1, user2, threshold):
common_items = extractor.get_common_items(self.user_dictionary, user1,
user2)
if not common_items:
return None
filtered_items = {}
for item in common_items:
context1 = self.user_dictionary[user1].item_contexts[item]
context2 = self.user_dictionary[user2].item_contexts[item]
context_similarity = context_utils.get_context_similarity(
context1, context2, self.context_rich_topics)
if context_similarity > threshold:
filtered_items[item] = context_similarity
numerator = 0
denominator1 = 0
denominator2 = 0
denominator3 = 0
user1_average = self.user_dictionary[user1].average_overall_rating
user2_average = self.user_dictionary[user2].average_overall_rating
for item in filtered_items.keys():
context_similarity = filtered_items[item]
user1_rating = self.user_dictionary[user1].item_ratings[item]
user2_rating = self.user_dictionary[user2].item_ratings[item]
numerator +=\
(user1_rating - user1_average) *\
(user2_rating - user2_average) *\
context_similarity
denominator1 += (user1_rating - user1_average)**2
denominator2 += (user2_rating - user2_average)**2
denominator3 += context_similarity**2
denominator = math.sqrt(denominator1 * denominator2 * denominator3)
if denominator == 0:
return None
return numerator / denominator
def calculate_users_similarity(self, user_dictionary, user1, user2):
common_items = extractor.get_common_items(user_dictionary, user1, user2)
if not common_items:
return None
if self._min_common_items is not None and len(
common_items) < self._min_common_items:
return None
user1_ratings =\
extractor.get_user_ratings(user_dictionary, user1, common_items)
user2_ratings =\
extractor.get_user_ratings(user_dictionary, user2, common_items)
similarity_value = similarity_calculator.calculate_similarity(
user1_ratings, user2_ratings, self._similarity_metric)
return similarity_value
def calculate_users_similarity(self, user_dictionary, user1, user2):
common_items = extractor.get_common_items(user_dictionary, user1,
user2)
if not common_items:
return None
if self._min_common_items is not None and len(
common_items) < self._min_common_items:
return None
user1_overall_ratings =\
extractor.get_user_ratings(user_dictionary, user1, common_items)
user1_multi_ratings =\
extractor.get_user_multi_ratings(user_dictionary, user1, common_items)
user2_overall_ratings =\
extractor.get_user_ratings(user_dictionary, user2, common_items)
user2_multi_ratings =\
extractor.get_user_multi_ratings(user_dictionary, user2, common_items)
num_criteria = len(user1_multi_ratings[0])
total_similarity = 0.
for i in xrange(0, num_criteria):
user1_criterion_item_ratings =\
extractor.get_matrix_column(user1_multi_ratings, i)
user2_criterion_item_ratings =\
extractor.get_matrix_column(user2_multi_ratings, i)
total_similarity += similarity_calculator.calculate_similarity(
user1_criterion_item_ratings, user2_criterion_item_ratings,
self._similarity_metric)
# We also add the overall similarity
total_similarity += similarity_calculator.calculate_similarity(
user1_overall_ratings, user2_overall_ratings,
self._similarity_metric)
average_similarity = total_similarity / (num_criteria + 1)
return average_similarity
def calculate_users_similarity(self, user_dictionary, user1, user2):
common_items = extractor.get_common_items(user_dictionary, user1, user2)
if not common_items:
return None
if self._min_common_items is not None and len(
common_items) < self._min_common_items:
return None
user1_overall_ratings =\
extractor.get_user_ratings(user_dictionary, user1, common_items)
user1_multi_ratings =\
extractor.get_user_multi_ratings(user_dictionary, user1, common_items)
user2_overall_ratings =\
extractor.get_user_ratings(user_dictionary, user2, common_items)
user2_multi_ratings =\
extractor.get_user_multi_ratings(user_dictionary, user2, common_items)
num_criteria = len(user1_multi_ratings[0])
total_similarity = 0.
for i in xrange(0, num_criteria):
user1_criterion_item_ratings =\
extractor.get_matrix_column(user1_multi_ratings, i)
user2_criterion_item_ratings =\
extractor.get_matrix_column(user2_multi_ratings, i)
total_similarity += similarity_calculator.calculate_similarity(
user1_criterion_item_ratings, user2_criterion_item_ratings,
self._similarity_metric)
# We also add the overall similarity
total_similarity += similarity_calculator.calculate_similarity(
user1_overall_ratings, user2_overall_ratings, self._similarity_metric)
average_similarity = total_similarity / (num_criteria + 1)
return average_similarity
def calculate_user_similarity(self, user_id1, user_id2, threshold):
common_items =\
extractor.get_common_items(self.user_dictionary, user_id1, user_id2)
if not common_items:
return None
numerator = 0
denominator1 = 0
denominator2 = 0
for item in common_items:
user1_rating = self.user_dictionary[user_id1].item_ratings[item]
user2_rating = self.user_dictionary[user_id2].item_ratings[item]
numerator += user1_rating * user2_rating
denominator1 += user1_rating**2
denominator2 += user2_rating**2
denominator = math.sqrt(denominator1) * math.sqrt(denominator2)
return numerator / denominator
def calculate_user_similarity(self, user_id1, user_id2, threshold):
common_items =\
extractor.get_common_items(self.user_dictionary, user_id1, user_id2)
if not common_items:
return None
numerator = 0
denominator1 = 0
denominator2 = 0
for item in common_items:
user1_rating = self.user_dictionary[user_id1].item_ratings[item]
user2_rating = self.user_dictionary[user_id2].item_ratings[item]
numerator += user1_rating * user2_rating
denominator1 += user1_rating ** 2
denominator2 += user2_rating ** 2
denominator = math.sqrt(denominator1) * math.sqrt(denominator2)
return numerator / denominator
