def year_week(y, w):
try:
if w not in ['', np.nan, None] and y not in ['', np.nan, None]:
week = Week(int(y), int(w))
date = week.startdate() + timedelta(3)
return date
else:
return ''
except:
return ''
def _parse_datetimes(date_int: int,
time_type: str,
date_format: str = "%Y%m%d") -> Union[pd.Timestamp]: # annotating nan errors
"""Convert a date or epiweeks string into timestamp objects.
Datetimes (length 8) are converted to their corresponding date, while epiweeks (length 6)
are converted to the date of the start of the week. Returns nan otherwise
Epiweeks use the CDC format.
:param date_int: Int representation of date.
:param date_format: String of the date format to parse.
:returns: Timestamp.
"""
date_str = str(date_int)
if time_type == "day":
return pd.to_datetime(date_str, format=date_format)
if time_type == "week":
epiwk = Week(int(date_str[:4]), int(date_str[-2:]))
return pd.to_datetime(epiwk.startdate())
return np.nan
n_neurons = n_neurons
data_path = data_path
ind_og = pd.read_csv(data_path)
# ind = data.set_index(["WEEK"])
region = "Region"
ind = ind_og.loc[ind_og["REGION"] == "Region " + str(r)]
ind_next = ind_og[0:300]
from epiweeks import Week, Year
date = []
for item, row in ind.iterrows():
week = Week(row["YEAR"], row["WEEK"])
# print(row["YEAR"], row["WEEK"], week.startdate())
date.append(week.startdate())
ind["DATE"] = date
new_ind = ind[['DATE', '% WEIGHTED ILI']] # ,'WEEK']]
new_ind.index = new_ind['DATE']
series = new_ind.drop(['DATE'], axis=1)
all_32_forecasts = []
for r in range(1, 11):
# date-time parsing function for loading the dataset
def parser(x):
return datetime.strptime('190' + x, '%Y-%m')
# convert time series into supervised learning problem
