def parse_train_data(self, filepath_or_buffer):
self._list_of_instances, self._list_of_labels = self._load_train_data(
filepath_or_buffer)
checks.check_equality(len(self._list_of_instances),
len(self._list_of_labels),
message="Instances of read data are not equal "
"to their.")
checks.check_equality(self.to_final_label(
self.to_interim_label([24, 42, 42])), [24, 42, 42],
message="Processing data methods are "
"not mutually inverse.")
self._list_of_samples = list(
map(self._to_sample, self._list_of_instances))
if self._debug:
print(len(self._list_of_instances))
print(self._list_of_instances[:3])
print(self._list_of_labels[:3])
print(self._list_of_samples[:3])
self._train_samples_num = int(self._proportion *
len(self._list_of_labels))
self._chknums = self._chknums[self._train_samples_num:]
self._answers_for_train = [
sorted(x) for x in self._list_of_labels[self._train_samples_num:]
]
def train(self, train_samples, train_labels, **kwargs):
checks.check_equality(len(train_samples),
len(train_labels),
message="Samples and labels have different "
"sizes")
self.most_popular_goods = kwargs["most_popular_goods"]
self.most_popular_good_ids = kwargs["most_popular_good_ids"]
checks.check_value(len(self.most_popular_good_ids),
lower=self.num_popular_ids,
strict_less=False,
var_name="most_popular_good_ids")
self.max_good_id = kwargs["max_good_id"]
# Get person ids from train samples, samples format:
# [[person_id, month, day], [person_id, month, day], ...].
persons_ids = [person_data[0] for person_data in train_samples]
for persons_id, label in zip(persons_ids, train_labels):
if self.orders.get(persons_id) is None:
self.orders[persons_id] = np.array(label)
else:
self.orders[persons_id] += np.array(label)
self.process_orders()
def train(self, train_samples, train_labels, **kwargs):
checks.check_equality(len(train_samples),
len(train_labels),
message="Samples and labels have different "
"sizes")
self.most_popular_goods = kwargs["most_popular_goods"]
# Get person ids from train samples, samples format:
# [[person_id, month, day], [person_id, month, day], ...].
persons_ids = [person_data[0] for person_data in train_samples]
self.latest_orders = dict(zip(persons_ids, train_labels))
def train(self, train_samples, train_labels, **kwargs):
"""
Train current model.
:param train_samples: array-like, sparse matrix.
Training data.
:param train_labels: array-like, sparse matrix.
Target values. Will be cast to train_samples’s dtype if necessary.
:param kwargs: dict, optional(default={}).
Additional keyword arguments.
"""
checks.check_equality(len(train_samples),
len(train_labels),
message="Samples and labels have different "
"sizes")
self.model.fit(train_samples, train_labels, **kwargs)
def fit(self, train_samples, train_labels, **kwargs):
checks.check_equality(len(train_samples),
len(train_labels),
message="Samples and labels have different "
"sizes")
persons_ids = [person_data[0] for person_data in train_samples]
for persons_id, label in zip(persons_ids, train_labels):
if self.orders.get(persons_id) is None:
self.orders[persons_id] = np.array(label)
else:
self.orders[persons_id] += np.array(label)
self.clustering_table = pd.DataFrame(self.model.fit_predict(
pd.DataFrame.from_dict(self.orders, orient="index")),
columns=[self.COL_NAME])
cluster_id = self.clustering_table[self.COL_NAME]\
.value_counts().index[0]
larg_clust_center = self.model.cluster_centers_[cluster_id]
self.largest_cluster_goods = (larg_clust_center >=
self.CLUSTER_BORDER).astype(np.int)
def train(self, train_samples, train_labels, **kwargs):
checks.check_equality(len(train_samples),
len(train_labels),
message="Samples and labels have different "
"sizes")
self.most_popular_goods = kwargs["most_popular_goods"]