def test_of_user_data():
session = DBManager.get_session()
user_list = session.query(User).all()
assert len(user_list) == 200
def test_cb_recommend():
''' Everytime client ask for recommendation will invoke this method, return a python object '''
db_path = "./test.sqlite"
# If db created then connect, otherwise, load test data.
if os.path.exists(db_path):
DBManager.init_db(db_name='test', is_echo=False)
else:
DBManager.init_db(db_name='test', is_echo=False)
Loader.load(file_path='./tests/food.json')
Loader.load(file_path='./tests/user.json')
Loader.load(file_path='./tests/review.json')
user_id = 20
session = DBManager.get_session()
food = pd.read_sql(session.query(Food).statement,
session.bind).rename(columns={'id': 'food_id'})
review = pd.read_sql(
session.query(UserRecommendationReview).filter(
user_id == UserRecommendationReview.user_id).statement,
session.bind)
review['is_accept'] = np.where(review['is_accept'] == True, 1, 0)
review = review.groupby([
'food_id'
])['is_accept'].mean().reset_index().sort_values(by=["is_accept"],
ascending=False)
final_recommendation = _recommend_food(review, food)
print(final_recommendation)
def recommend(
self,
user=None,
*args,
**kwargs
) -> 'Food[] , high recommendation prority in lower index ':
''' Everytime client ask for recommendation will invoke this method, return a python object '''
user_id = user.id
session = DBManager.get_session()
food = pd.read_sql(session.query(Food).statement,
session.bind).rename(columns={'id': 'food_id'})
review = pd.read_sql(
session.query(UserRecommendationReview).filter(
user_id == UserRecommendationReview.user_id).statement,
session.bind)
review['is_accept'] = np.where(review['is_accept'] == True, 1, 0)
review = review.groupby([
'food_id'
])['is_accept'].mean().reset_index().sort_values(by=["is_accept"],
ascending=False)
final_recommendation = self._recommend_food(review, food)
food_list = self._id_to_food_list(session, final_recommendation)
return food_list
def test_of_food_data():
session = DBManager.get_session()
food_list = session.query(Food).order_by(Food.id.asc()).all()
assert len(food_list) == 295
def recommend(self, max_lenght=10, *args, **kwargs):
session = DBManager.get_session()
avaiable_foods = session.query(Food).all()
random.seed(random.randint(0, 999))
if not avaiable_foods: return list()
return [random.choice(avaiable_foods) for _ in range(max_lenght)]
def recommend(self, user=None):
if self.first_time:
self.first_time = False
try:
self.requirement = user.basal_metabolic_rate
except:
pass
self.min_cal, self.max_cal = self.requirement * 0.15, self.requirement * 0.6
self.session = DBManager.get_session()
self.foodlist = self.session.query(Food).filter(
self.min_cal < Food.calories).filter(
Food.calories < self.max_cal).order_by(
Food.calories.asc()).all()
self.target_cal = self.requirement * 4.8
else:
# update a_dict_take_or_not, 第二次以後被呼叫才會計算
Last_review = self.session.query(
UserRecommendationReview).order_by(
UserRecommendationReview.created_datetime.desc()).first()
food_taken_id = Last_review.food_id
for food in self.list:
if food.id not in self.a_dict_take_or_not:
self.a_dict_take_or_not[food.id] = [0, 0]
self.a_dict_take_or_not[food.id][0] += 1
if food.id == food_taken_id:
cal = food.calories
self.a_dict_take_or_not[food.id][1] += 1
break
# update
last_requirement = self.cal_requirement(self.time - 1)
try:
self.target_cal += cal - last_requirement
except:
self.target_cal += self.session.query(Food).filter(
Food.id ==
food_taken_id).first().calories - last_requirement
# 開始處理
requirement = self.cal_requirement(self.time)
this_meal_target = int(self.target_cal / self.meals_left)
r_list = self.recommend_food_with_possibility(this_meal_target,
requirement,
self.foodlist)
self.meals_left -= 1
self.time += 1
self.list = r_list
return r_list
def test_of_food_data():
session = DBManager.get_session()
food_list = session.query(Food).order_by(Food.id.asc()).all()
assert len(food_list) == 295
user_list = session.query(User).all()
assert len(user_list) == 201
review_list = session.query(UserRecommendationReview).all()
assert len(review_list) > 0
def prepare_recommedation_foods(self):
try:
self.requirement = user.basal_metabolic_rate
except:
pass
self.min_cal, self.max_cal = self.requirement * 0.15, self.requirement * 0.6
self.session = DBManager.get_session()
self.foodlist = self.session.query(Food).filter(
self.min_cal < Food.calories).filter(
Food.calories < self.max_cal).order_by(
Food.calories.asc()).all()
self.lose_one_kg_scedule()
def _login(self):
session = DBManager.get_session()
while True:
print('\nRandom User List')
for user in random.choices(session.query(User).all(), k=10):
print(user)
username = Cmd.get_input('Please enter user name to login')
user = session.query(User).filter(
User.name == username).one_or_none()
if user is None:
print('User not found !')
else:
self.application.login(user)
break
def test_recommend():
''' Everytime client ask for recommendation will invoke this method, return a python object '''
db_path = "./test.sqlite"
# If db created then connect, otherwise, load test data.
if os.path.exists(db_path):
DBManager.init_db(db_name='test', is_echo=False)
else:
DBManager.init_db(db_name='test', is_echo=False)
Loader.load(file_path='./tests/food.json')
Loader.load(file_path='./tests/user.json')
Loader.load(file_path='./tests/review.json')
# Setup session and tables
session = DBManager.get_session()
food = pd.read_sql(session.query(Food).statement,session.bind).rename(columns={'id': 'food_id'})
user = pd.read_sql(session.query(User).statement,session.bind).rename(columns={'id': 'user_id'})
review = pd.read_sql(session.query(UserRecommendationReview).statement,session.bind)
# Data processing & prepare
review['is_accept'] = np.where(review['is_accept']==True, 1, -1)
review = review.groupby(['user_id','food_id'])['is_accept'].sum().reset_index()
rating_pivot = review.pivot(index='user_id',columns = 'food_id',values='is_accept').fillna(0)
rating_df = rating_pivot.as_matrix()
user_ratings_mean = np.mean(rating_df, axis = 1)
r_demeaned = rating_df - user_ratings_mean.reshape(-1, 1)
# Matrix factorization
U, sigma, Vt = svds(r_demeaned, k = 50)
sigma = np.diag(sigma)
prediction_all= np.dot(np.dot(U, sigma), Vt) + user_ratings_mean.reshape(-1, 1)
prediction_df = pd.DataFrame(prediction_all, columns = rating_pivot.columns)
# Recommend top 10 food that the user hasn't been recommended to
recommendation_list = recommend_food(prediction_df, 130, food, review, 10)
print(recommendation_list)
result = session.query(Food).filter(Food.id.in_(recommendation_list)).all()
food_map = {food.id: food for food in result}
food_list = [food_map[id] for id in recommendation_list]
for food in food_list:
food.show_detail()
def _find_similar_food(self, food_id, food, positive=1) -> 'food.id[]':
try:
count = CountVectorizer()
count_matrix = count.fit_transform(food['name'])
cosine_sim = cosine_similarity(count_matrix, count_matrix)
if positive == 1:
score_series = pd.Series(
cosine_sim[food_id]).sort_values(ascending=False)
else: # finds the least similar
score_series = pd.Series(
cosine_sim[food_id]).sort_values(ascending=True)
similar_food = list(score_series.iloc[1:11].index)
return similar_food
except Exception as e:
session = DBManager.get_session()
random_food = random.choice(session.query(Food).all())
return list([random_food.id])
def recommend(
self,
user=None,
*args,
**kwargs
) -> 'Food[] , high recommendation prority in lower index ':
''' Everytime client ask for recommendation will invoke this method, return a python object '''
user_id = user.id
session = DBManager.get_session()
food = pd.read_sql(session.query(Food).statement,
session.bind).rename(columns={'id': 'food_id'})
user = pd.read_sql(session.query(User).statement,
session.bind).rename(columns={'id': 'user_id'})
review = pd.read_sql(
session.query(UserRecommendationReview).statement, session.bind)
# Data processing & prepare
review['is_accept'] = np.where(review['is_accept'] == True, 1, -1)
review = review.groupby(['user_id',
'food_id'])['is_accept'].sum().reset_index()
rating_pivot = review.pivot(index='user_id',
columns='food_id',
values='is_accept').fillna(0)
# Turn review to ratings matrix
rating_df = rating_pivot.as_matrix()
user_ratings_mean = np.mean(rating_df, axis=1)
r_demeaned = rating_df - user_ratings_mean.reshape(-1, 1)
# Matrix factorization
U, sigma, Vt = svds(r_demeaned, k=10)
sigma = np.diag(sigma)
prediction_all = np.dot(np.dot(U, sigma),
Vt) + user_ratings_mean.reshape(-1, 1)
prediction_df = pd.DataFrame(prediction_all,
columns=rating_pivot.columns)
recommendation_list = self._recommend_for_user(prediction_df, user_id,
food, review)
food_list = self._id_to_food_list(session, recommendation_list)
return food_list
def recommend(self, user=None, *args, **kwargs):
if self.first_time:
self.first_time = False
try:
self.requirement = user.basal_metabolic_rate
except:
pass
self.min_cal, self.max_cal = self.requirement * 0.15, self.requirement * 0.6
self.session = DBManager.get_session()
self.foodlist = self.session.query(Food).filter(
self.min_cal < Food.calories).filter(
Food.calories < self.max_cal).order_by(
Food.calories.asc()).all()
self.lose_one_kg_scedule()
r_f = self.path.pop(0)
return [r_f]
def food_name_pattern_dict(self) -> 'dict(pattern : count_time)':
'''
Put all element of Food.name in a dict (self._pattern)
and the value is the referenced time
'''
if any(self._pattern): return self._pattern
session = DBManager.get_session()
foods = session.query(Food).all()
for food in foods:
name_seq = food.name.lower().split(' ')
for pattern in name_seq:
if pattern in [
'!', '@', '#', '$', '%', '^', '&', '*', '(', ')'
]:
continue
if pattern in self._pattern: self._pattern[pattern] += 1
else:
self._pattern[pattern] = 1
return self._pattern
def test_proxy_run():
Loader.load(file_path=os.path.join(os.getcwd(), 'data', 'user.json'))
Loader.load(file_path=os.path.join(os.getcwd(), 'data', 'food.json'))
Loader.load(file_path=os.path.join(os.getcwd(), 'data', 'review.json'))
session = DBManager.get_session()
foods = session.query(Food).all()
random_user_name = ''.join(
[random.choice(string.ascii_lowercase) for _ in range(10)])
user = User(name=random_user_name, weight=70, height=170).save()
application = Application(db_is_echo=False)
user_proxy = UserProxyCollection.default_proxy(
user=user,
diet_schedule=DietScheduleCollection.defautl_schedule,
application=application)
pickup_food_times = 30
for _ in range(pickup_food_times):
user_proxy.choice_food()
user_proxy.report()
def __init__(self, db_name='db', *args, **kwrags):
if db_name is None or not isinstance(db_name, str): raise Exception('please provide DB name to init service')
self.db_name = db_name
DBManager.init_db(db_name=db_name, is_echo=kwrags.get('db_is_echo', True))
self.session = DBManager.get_session()