def test_holidays_for_year(self):
d = SchoolHolidayDates()
res = d.holidays_for_year(2018)
self.assertEquals(len(res), 151)
for k, v in res.items():
self.assertEquals(sorted(v.keys()), self.EXPECTED_KEYS)
with self.assertRaisesRegexp(UnsupportedYearException,
"No data for year: 2024"):
self.assertEquals({}, d.holidays_for_year(2024))
def test_holidays_for_year(self):
d = SchoolHolidayDates()
res = d.holidays_for_year(2018)
self.assertEquals(len(res), 151)
for k, v in res.items():
self.assertEquals(sorted(v.keys()), self.EXPECTED_KEYS)
with self.assertRaisesRegexp(
UnsupportedYearException, "No data for year: 2021"
):
self.assertEquals({}, d.holidays_for_year(2021))
def test_supported_holidays_are_complete(self):
d = SchoolHolidayDates()
res = d.holidays_for_year(2019)
names = set()
for _, v in res.items():
names.add(v["nom_vacances"])
expected = set(SchoolHolidayDates.SUPPORTED_HOLIDAY_NAMES)
self.assertEquals(names, expected)
def test_supported_holidays_are_complete(self):
d = SchoolHolidayDates()
res = d.holidays_for_year(2019)
names = set()
for _, v in res.items():
names.add(v["nom_vacances"])
expected = set(SchoolHolidayDates.SUPPORTED_HOLIDAY_NAMES)
self.assertEquals(names, expected)
def get_encoded_calendar_df(data_final):
# Load school holidays for France
fr_holidays = SchoolHolidayDates()
df_vacances = pd.DataFrame()
for year in list(set(data_final['year'])):
df_vacances = pd.concat([df_vacances, pd.DataFrame.from_dict(fr_holidays.holidays_for_year(year)).T])
# Load bank holidays for France
df_jf = pd.DataFrame()
for year in list(set(data_final['year'])):
df_jf = pd.concat([df_jf, pd.DataFrame([
{'date': el[0], 'jour_ferie': el[1]} for el in sorted(holidays.FRA(years=year).items())])])
# Merge school and bank holidays
df_holidays = pd.merge(df_vacances, df_jf, how='outer', on='date')
# Create features from df_holidays dataframes (school holidays and bank holidays):
# - 3 binary features for school holidays, taking 1 if the given zone is on holiday, else 0 (vacances_zone_a,
# vacances_zone_b, vacances_zone_c)
# Definition of a dictionary to encode boolean into numeric
dict_map_vac = {
True: 1,
False: 0
}
# Apply dictionary to each holiday column for the three zones (A, B, C)
df_holidays['vacances_zone_a'] = df_holidays['vacances_zone_a'].map(dict_map_vac)
df_holidays['vacances_zone_b'] = df_holidays['vacances_zone_b'].map(dict_map_vac)
df_holidays['vacances_zone_c'] = df_holidays['vacances_zone_c'].map(dict_map_vac)
# - 1 binary feature for bank holiday, taking 1 if it is a bank holiday, else 0
# The column "jour ferie" contains either the name of the holiday or a missing value (NaN)
# The idea is to put a '1' when it's a holiday (i.e. when the value is different from nan, else 0)
df_holidays['jour_ferie'] = df_holidays['jour_ferie'].map(lambda x: 1 if str(x) != 'nan' else 0)
# - To go further: Try to create a combined feature with school and bank holidays
df_holidays['holiday'] = df_holidays['vacances_zone_a'] + df_holidays['vacances_zone_b'] + df_holidays[
'vacances_zone_c'] + df_holidays['jour_ferie']
df_holidays['date'] = df_holidays['date'].map(lambda x: str(x))
data_final_cal = pd.merge(data_final, df_holidays, how='left', left_on='release_date', right_on='date').fillna(0)
data_final_cal['month'] = data_final_cal['release_date'].map(lambda x: int(x[5:7]))
data_final_cal = apply_cos(data_final_cal, 'month', 'cos_month', 12)
return data_final_cal