tomonori-masui

ts_al_2diff = (ts_al_diff - ts_al_diff.shift(1)).dropna()
tsplot(ts_al_2diff)

tomonori-masui

ts_al_diff = (ts_al - ts_al.shift(12)).dropna()
tsplot(ts_al_diff)

tomonori-masui

from sktime.datasets import load_airline
ts_al = load_airline()

tsplot(ts_al)

tomonori-masui

from sktime.forecasting.compose import TransformedTargetForecaster

def create_forecaster_w_detrender(degree=1):

    # creating forecaster with LightGBM
    regressor = lgb.LGBMRegressor()
    forecaster = TransformedTargetForecaster(
        [
            ("detrend", Detrender(forecaster=PolynomialTrendForecaster(degree=degree))),
            (

tomonori-masui

from sktime.forecasting.trend import PolynomialTrendForecaster
from sktime.transformations.series.detrend import Detrender

# linear detrending
forecaster = PolynomialTrendForecaster(degree=1)
transformer = Detrender(forecaster=forecaster)
yt = transformer.fit_transform(wpi_train)

forecaster = PolynomialTrendForecaster(degree=1)
fh_ins = -np.arange(len(wpi_train))  # in-sample forecasting horizon

tomonori-masui

param_grid = {"window_length": [5, 10, 15, 20, 25, 30]}
forecaster = create_forecaster()
wpi_lgb_mae, wpi_lgb_mape = grid_serch_forecaster(
    wpi_train, wpi_test, forecaster, param_grid
)

tomonori-masui

fh = np.arange(test_len) + 1
forecast = forecaster.predict(fh=fh)
forecast_int = forecaster.predict_interval(fh=fh, coverage=coverage)['Coverage'][coverage]
wpi_arima_mae, wpi_arima_mape = plot_forecast(
    wpi_train, wpi_test, forecast, forecast_int
)

tomonori-masui

ts_wpi.index = ts_wpi.index.to_period("Q")
test_len = int(len(ts_wpi) * 0.25)
wpi_train, wpi_test = ts_wpi.iloc[:-test_len], ts_wpi.iloc[-test_len:]

forecaster = AutoARIMA(suppress_warnings=True)
forecaster.fit(wpi_train)
forecaster.summary()

tomonori-masui

ts_wpi_2diff = (ts_wpi_diff - ts_wpi_diff.shift(1)).dropna()
tsplot(ts_wpi_2diff)

tomonori-masui

ts_wpi_diff = (ts_wpi - ts_wpi.shift(1)).dropna()
tsplot(ts_wpi_diff)