class SimilarProductsUsers:
def __init__(self):
self.usersDao = Users()
self.user_embeddings = self.usersDao.get_embeddings()
self.productsDao = Products()
self.product_embeddings = self.productsDao.get_embeddings()
def neighbors_products(self,
user_id,
list_current_product_id,
n_closest=10):
user_idx = self.usersDao.user_id_to_idx(user_id)
dists = np.dot(self.product_embeddings, self.user_embeddings[user_idx])
closest_product_idx = np.argsort(dists)[-n_closest:]
similarity_dict = {}
for c in closest_product_idx:
local_product_id = self.productsDao.idx_to_products(
[c])[['ProdutoId']].values[0][0]
if local_product_id not in list_current_product_id:
similarity_dict.update({local_product_id: dists[c]})
return similarity_dict
def neighbors_user_idx(self, product_idx, n_closest=5):
dists = np.dot(self.user_embeddings,
self.product_embeddings[product_idx])
closest_user_idx = np.argsort(dists)[-n_closest:]
return closest_user_idx
class SimilarUsers:
def __init__(self):
self.usersDao = Users()
self.user_embeddings = self.usersDao.get_embeddings()
def neighbors_user_id(self, user_id, n_closest=10):
user_idx = self.usersDao.user_id_to_idx(user_id)
dists = np.dot(self.user_embeddings, self.user_embeddings[user_idx])
closest_user_idx = np.argsort(dists)[-n_closest:]
similarity_dict = {}
for c in closest_user_idx:
local_user_id = self.usersDao.idx_to_users([c])[['userId'
]].values[0][0]
if local_user_id not in [user_id]:
similarity_dict.update({local_user_id: dists[c]})
return similarity_dict
class QuantityProductRegression:
def __init__(self):
self.usersDao = Users()
self.productDao = Products()
def fit_transform(self, user_id, df_products):
list_product_id = list(df_products.ProdutoId.values)
user_embedding = self.usersDao.user_id_to_embedding([user_id])
product_embeddings = self.productDao.product_id_to_embedding(
list_product_id)
df_dot_product = normalized_dot_product(user_embedding,
product_embeddings)
df_regression = rfr.predict(df_dot_product)
dict_product_regression = {}
for _, row in df_regression.iterrows():
dict_product_regression.update(
{row['ProdutoId']: row['ProductQuantity']})
df_products['ProductQuantity'] = df_products['ProdutoId'].map(
dict_product_regression)
return df_products
class SimilarityEmbeddings:
def __init__(self):
self.usersDao = Users()
self.productDao = Products()
def cossine_distance(self, list_user_id, list_product_id, n_closest=6):
df_user_embedding = self.usersDao.user_id_to_embedding(list_user_id)
user_embedding = df_user_embedding.iloc[:, 1:].values
df_product_embeddings = self.productDao.product_id_to_embedding(
list_product_id)
df_product_embeddings.reset_index(inplace=True)
product_embeddings = df_product_embeddings.iloc[:, 2:].values
dists = np.dot(product_embeddings, user_embedding.reshape(-1, 1))
closest_idx = np.argsort(dists.reshape(1, -1)[0])[-n_closest:]
similarity_dict = {}
for c in closest_idx:
local_id = df_product_embeddings.iloc[c, :][['ProdutoId'
]].values[0]
similarity_dict.update({local_id: dists[c][0]})
return similarity_dict
def __init__(self):
self.usersDao = Users()
self.productDao = Products()
def __init__(self):
self.usersDao = Users()
self.user_embeddings = self.usersDao.get_embeddings()