def __init__(self,
media_type='movie',
db_name='VionelDB',
collection_name='BoxerMovies',
hostname='192.168.1.80',
port=27017):
if media_type == 'movie':
self.feature_weight_dict = {
'imdbDirector': 0.7,
'imdbGenre': 0.7 * 69325.4760905 / 223399.25309 / 3,
'imdbKeyword': 0.5,
'wikiKeyword': 0,
'vionelTheme': 0.5,
'vionelScene': 0.35,
'locationCountry': 0.3,
'locationCity': 0.5,
'imdbMainactor': 0.7 * 215719.641732 / 223399.25309 / 3,
'RGB': 0.25,
'brightness': 0.25
}
elif media_type == 'tv':
self.feature_weight_dict = {
'imdbCreator': 1,
'imdbGenre': 1,
'imdbMainactor': 1,
'imdbKeyword': 10
}
self.mongo_manager = MongoManager(db_name, collection_name, hostname,
port)
self.media_type = media_type
def __init__(self, media_type='movie',
db_name='VionelDB',
collection_name='BoxerMovies',
hostname='192.168.1.80',
port=27017):
if media_type == 'movie':
self.feature_weight_dict = {
'imdbDirector': 0.7,
'imdbGenre': 0.7 * 69325.4760905 / 223399.25309 / 3,
'imdbKeyword': 0.5,
'wikiKeyword': 0,
'vionelTheme': 0.5,
'vionelScene': 0.35,
'locationCountry': 0.3,
'locationCity': 0.5,
'imdbMainactor': 0.7 * 215719.641732/223399.25309 / 3,
'RGB': 0.25,
'brightness': 0.25
}
elif media_type == 'tv':
self.feature_weight_dict = {
'imdbCreator': 1,
'imdbGenre': 1,
'imdbMainactor': 1,
'imdbKeyword': 10
}
self.mongo_manager = MongoManager(db_name, collection_name, hostname, port)
self.media_type = media_type
def __init__(self):
self.feature_weight_dict = {
'imdbDirector': 0.7,
'imdbGenre': 0.5,
'imdbKeyword': 0.6,
'wikiKeyword': 1.3,
'vionelTheme': 1.3,
'vionelScene': 0.35,
'locationCountry': 0.3,
'locationCity': 0.5,
'imdbMainactor': 0.9,
'RGB': 0.25,
'brightness': 0.25
}
self.mongo_manager = MongoManager('VionelMovies', 'BoxerMovies', '172.17.42.1', 27017, '', '')
class SimilarityRecommender(object):
def __init__(self):
self.feature_weight_dict = {
'imdbDirector': 0.7,
'imdbGenre': 0.5,
'imdbKeyword': 0.6,
'wikiKeyword': 1.3,
'vionelTheme': 1.3,
'vionelScene': 0.35,
'locationCountry': 0.3,
'locationCity': 0.5,
'imdbMainactor': 0.9,
'RGB': 0.25,
'brightness': 0.25
}
self.mongo_manager = MongoManager('VionelMovies', 'BoxerMovies', '172.17.42.1', 27017, '', '')
def __get_imdbid_feature_dict(self, feature_name):
result_dict = {}
all_movies_feature_dict_list = self.mongo_manager.exec_query({}, {"imdbId": 1, feature_name: 1, "_id": 0})
for movie in all_movies_feature_dict_list:
imdbid = movie["imdbId"]
try:
feature = movie[feature_name]
result_dict[imdbid] = feature
except KeyError:
continue
return result_dict
def __get_imdbid_similarity_dict(self, movieid_list, recommended_by):
movieid_with_featureid_dict = {}
input_featureid_with_number_dict = {}
movieid_with_featureid_dict = self.__get_imdbid_feature_dict(recommended_by)
result_dict = {}
if recommended_by == "imdbDirectors" or recommended_by == "imdbGenres" or recommended_by == "locationCountry" or recommended_by == "locationCity" or recommended_by == "vionelScene" or recommended_by == "imdbMainactors" or recommended_by == "RGB" or recommended_by == "Brightness":
input_featureid_with_number_dict = intersection_of_values_for_certain_keys(movieid_list, movieid_with_featureid_dict)
all_featureid_list = input_featureid_with_number_dict.keys()
for k, v in movieid_with_featureid_dict.items():
intersection_list = list(set(v).intersection(set(all_featureid_list)))
if not intersection_list:
result_dict[k] = 0
continue
compared_movie_feature_num_dict = intersection_of_values_for_certain_keys([k], movieid_with_featureid_dict)
cosine_score = calculate_cosine(input_featureid_with_number_dict, compared_movie_feature_num_dict)
result_dict[k] = cosine_score
return result_dict
else:
input_movie_features = []
input_movie_features = union_of_values_for_spec_keys(movieid_list, movieid_with_featureid_dict)
coefficient = 0.1
for k, v in movieid_with_featureid_dict.items():
intersection_num = len(list(set(v).intersection(set(input_movie_features))))
score = intersection_num * coefficient
if score > 1:
score = 1
result_dict[k] = score
return result_dict
def __language_filter(self, input_movieid_list, combined_movieid_sim_counter):
imdbid_language_dict = self.__get_imdbid_feature_dict("language")
languages_in_liked_list = []
for item in input_movieid_list:
try:
languages_in_liked_list += imdbid_language_dict[item]
except KeyError:
continue
languages_in_liked_list = list(set(languages_in_liked_list))
delete_list = []
for imdbid in combined_movieid_sim_counter:
language_list = imdbid_language_dict[imdbid]
if not language_list:
continue
intersection_list = list(set(languages_in_liked_list).intersection(set(language_list)))
if not intersection_list: # 如果为空,则排除此电影
delete_list.append(imdbid)
for x in delete_list:
del combined_movieid_sim_counter[x]
return combined_movieid_sim_counter
def __multiply_coefficient(self, movieid_score_counter, coefficient):
result_count = Counter()
for k, v in movieid_score_counter.items():
result_count[k] = v * coefficient
return result_count
def features_contribute_most(self, recommended_movies_dict):
"""Get the features that contribute most for each recommended movie."""
movieid_featurescore_dict = {}
for movieid in recommended_movies_dict:
feature_score_dict = {}
for feature in self.feature_weight_dict:
variable_name = feature.lower()
exec "feature_score_dict['%s'] = self.%s_movieid_sim_counter['%s']" % (feature, variable_name, movieid)
movieid_featurescore_dict[movieid] = feature_score_dict
return movieid_featurescore_dict
def recommend_for_each_feature(self, input_movieid_list, num_of_recommended_movies):
"""Generate variables dynamicly
Generate 'self.imdbdirectors_movieid_sim_counter' this kind
of variables.
We will generate for each feature which will be used in the
following steps.
"""
for feature in self.feature_weight_dict:
feature_movieid_sim_dict = self.__get_imdbid_similarity_dict(input_movieid_list, feature)
feature_movieid_sim_counter = Counter(feature_movieid_sim_dict)
feature_movieid_sim_counter = self.__multiply_coefficient(feature_movieid_sim_counter, self.feature_weight_dict[feature])
for movieid in input_movieid_list:
del feature_movieid_sim_counter[movieid]
variable_name = feature.lower()
exec "self.%s_movieid_sim_counter = feature_movieid_sim_counter" % variable_name
def recommend(self, input_movieid_list, num_of_recommended_movies):
"""Return recommended movies and the features that contribute most in this recommendation.
Format of the return(if num_of_recommended_movies is 2):
{
"movie": {
"tt0340855": 0.6837561795878957,
"tt1124035": 0.9627459173643833,
},
"reason": {
"tt0340855": {
"imdbDirector": 0,
"brightness": 0.025,
"locationCountry": 0.21213203435596423,
"vionelTheme": 0.13,
"RGB": 0.0625,
"locationCity": 0.20412414523193148,
"wikiKeyword": 0,
"imdbGenre": 0.05,
"vionelScene": 0,
"imdbMainactor": 0,
"imdbKeyword": 0
},
"tt1124035": {
"imdbDirector": 0,
"brightness": 0,
"locationCountry": 0.21213203435596423,
"vionelTheme": 0,
"RGB": 0.0625,
"locationCity": 0.15811388300841897,
"wikiKeyword": 0.13,
"imdbGenre": 0.05,
"vionelScene": 0.35,
"imdbMainactor": 0,
"imdbKeyword": 0
}
}
"""
self.recommend_for_each_feature(input_movieid_list, num_of_recommended_movies)
combined_movieid_sim_counter = Counter()
for feature in self.feature_weight_dict:
exec "combined_movieid_sim_counter += self.%s_movieid_sim_counter" % feature.lower()
# filter
combined_movieid_sim_counter = self.__language_filter(input_movieid_list, combined_movieid_sim_counter)
final_recommended_movies_dict = dict(combined_movieid_sim_counter.most_common(num_of_recommended_movies))
movieid_featurewithscore_dict = self.features_contribute_most(final_recommended_movies_dict)
# print reason_tuple_list
result_dict = dict()
result_dict["movie"] = final_recommended_movies_dict
result_dict["reason"] = movieid_featurewithscore_dict
return result_dict
class SimilarityRecommender(object):
def __init__(self,
media_type='movie',
db_name='VionelDB',
collection_name='BoxerMovies',
hostname='192.168.1.80',
port=27017):
if media_type == 'movie':
self.feature_weight_dict = {
'imdbDirector': 0.7,
'imdbGenre': 0.7 * 69325.4760905 / 223399.25309 / 3,
'imdbKeyword': 0.5,
'wikiKeyword': 0,
'vionelTheme': 0.5,
'vionelScene': 0.35,
'locationCountry': 0.3,
'locationCity': 0.5,
'imdbMainactor': 0.7 * 215719.641732 / 223399.25309 / 3,
'RGB': 0.25,
'brightness': 0.25
}
elif media_type == 'tv':
self.feature_weight_dict = {
'imdbCreator': 1,
'imdbGenre': 1,
'imdbMainactor': 1,
'imdbKeyword': 10
}
self.mongo_manager = MongoManager(db_name, collection_name, hostname,
port)
self.media_type = media_type
def __del__(self):
self.mongo_manager.close()
def __get_imdbid_feature_dict(self, feature_name):
result_dict = {}
all_movies_feature_dict_list = self.mongo_manager.exec_query(
{}, {
"imdbId": 1,
feature_name: 1,
"_id": 0
})
for movie in all_movies_feature_dict_list:
imdbid = movie["imdbId"]
try:
feature = movie[feature_name]
result_dict[imdbid] = feature
except KeyError:
continue
return result_dict
def __get_imdbid_similarity_dict(self, movieid_list, recommended_by):
movieid_with_featureid_dict = {}
input_featureid_with_number_dict = {}
movieid_with_featureid_dict = self.__get_imdbid_feature_dict(
recommended_by)
result_dict = {}
if recommended_by == 'imdbMainactor' or recommended_by == 'imdbDirector' or recommended_by == 'imdbGenre' or recommended_by == "locationCountry" or recommended_by == "locationCity" or recommended_by == "vionelScene" or recommended_by == "RGB" or recommended_by == "brightness" or recommended_by == 'imdbCreator':
input_featureid_with_number_dict = intersection_of_values_for_certain_keys(
movieid_list, movieid_with_featureid_dict)
all_featureid_list = input_featureid_with_number_dict.keys()
for k, v in movieid_with_featureid_dict.items():
intersection_list = list(
set(v).intersection(set(all_featureid_list)))
if not intersection_list:
result_dict[k] = 0
continue
compared_movie_feature_num_dict = intersection_of_values_for_certain_keys(
[k], movieid_with_featureid_dict)
cosine_score = calculate_cosine(
input_featureid_with_number_dict,
compared_movie_feature_num_dict)
result_dict[k] = cosine_score
return result_dict
else:
input_movie_features = []
input_movie_features = union_of_values_for_spec_keys(
movieid_list, movieid_with_featureid_dict)
coefficient = 0.1
for k, v in movieid_with_featureid_dict.items():
intersection_num = len(
list(set(v).intersection(set(input_movie_features))))
score = intersection_num * coefficient
if score > 1:
score = 1
result_dict[k] = score
return result_dict
def __language_filter(self, input_movieid_list,
combined_movieid_sim_counter):
imdbid_language_dict = self.__get_imdbid_feature_dict("language")
languages_in_liked_list = []
for item in input_movieid_list:
try:
languages_in_liked_list += imdbid_language_dict[item]
except KeyError:
continue
languages_in_liked_list = list(set(languages_in_liked_list))
delete_list = []
for imdbid in combined_movieid_sim_counter:
language_list = imdbid_language_dict[imdbid]
if not language_list:
continue
intersection_list = list(
set(languages_in_liked_list).intersection(set(language_list)))
if not intersection_list: # 如果为空,则排除此电影
delete_list.append(imdbid)
for x in delete_list:
del combined_movieid_sim_counter[x]
return combined_movieid_sim_counter
def __multiply_coefficient(self, movieid_score_counter, coefficient):
result_count = Counter()
for k, v in movieid_score_counter.items():
result_count[k] = v * coefficient
return result_count
def features_contribute_most(self, recommended_movies_dict):
"""Get the features that contribute most for each recommended movie."""
movieid_featurescore_dict = {}
for movieid in recommended_movies_dict:
feature_score_dict = {}
for feature in self.feature_weight_dict:
variable_name = feature.lower()
exec "feature_score_dict['%s'] = self.%s_movieid_sim_counter['%s']" % (
feature, variable_name, movieid)
movieid_featurescore_dict[movieid] = OrderedDict(
sorted(feature_score_dict.items(),
key=lambda t: t[1],
reverse=True))
return movieid_featurescore_dict
def recommend_for_each_feature(self, input_movieid_list,
num_of_recommended_movies):
"""Generate variables dynamicly
Generate 'self.imdbdirectors_movieid_sim_counter' this kind
of variables.
We will generate for each feature which will be used in the
following steps.
"""
for feature in self.feature_weight_dict:
feature_movieid_sim_dict = self.__get_imdbid_similarity_dict(
input_movieid_list, feature)
feature_movieid_sim_counter = Counter(feature_movieid_sim_dict)
feature_movieid_sim_counter = self.__multiply_coefficient(
feature_movieid_sim_counter, self.feature_weight_dict[feature])
for movieid in input_movieid_list:
del feature_movieid_sim_counter[movieid]
variable_name = feature.lower()
exec "self.%s_movieid_sim_counter = feature_movieid_sim_counter" % variable_name
def recommend(self, input_movie, num_of_recommended_movies):
"""Return recommended movies and the features that contribute most in this recommendation.
Format of the return(if num_of_recommended_movies is 2):
{
"movie": {
"tt0340855": 0.6837561795878957,
"tt1124035": 0.9627459173643833,
},
"reason": {
"tt0340855": {
"imdbDirector": 0,
"brightness": 0.025,
"locationCountry": 0.21213203435596423,
"vionelTheme": 0.13,
"RGB": 0.0625,
"locationCity": 0.20412414523193148,
"wikiKeyword": 0,
"imdbGenre": 0.05,
"vionelScene": 0,
"imdbMainactor": 0,
"imdbKeyword": 0
},
"tt1124035": {
"imdbDirector": 0,
"brightness": 0,
"locationCountry": 0.21213203435596423,
"vionelTheme": 0,
"RGB": 0.0625,
"locationCity": 0.15811388300841897,
"wikiKeyword": 0.13,
"imdbGenre": 0.05,
"vionelScene": 0.35,
"imdbMainactor": 0,
"imdbKeyword": 0
}
}
"""
# First check if the id is existed.
assert self.mongo_manager.exec_query({'imdbId': input_movie}, {})
input_movieid_list = [input_movie]
self.recommend_for_each_feature(input_movieid_list,
num_of_recommended_movies)
combined_movieid_sim_counter = Counter()
for feature in self.feature_weight_dict:
exec "combined_movieid_sim_counter += self.%s_movieid_sim_counter" % feature.lower(
)
if self.media_type == 'movie':
# filter
combined_movieid_sim_counter = self.__language_filter(
input_movieid_list, combined_movieid_sim_counter)
final_recommended_movies_dict = dict(
combined_movieid_sim_counter.most_common(
num_of_recommended_movies))
final_ordered_recommended_movies_dict = OrderedDict(
sorted(final_recommended_movies_dict.items(),
key=lambda t: t[1],
reverse=True))
movieid_featurewithscore_dict = self.features_contribute_most(
final_recommended_movies_dict)
# print reason_tuple_list
result_dict = dict()
result_dict["movie"] = final_ordered_recommended_movies_dict
result_dict["reason"] = movieid_featurewithscore_dict
return result_dict
