sktime_forecasting_pipeline.py

#### Step 1: Load data and simulate missing value ----
import numpy as np
from sktime.datasets import load_airline
y = load_airline()
y[2] = np.nan

#### Step 2: Create pipeline with preprocessing ----
from sktime.forecasting.compose import ForecastingPipeline
from sktime.transformations.series.impute import Imputer
from sktime.transformations.series.boxcox import LogTransformer
from sktime.forecasting.compose import TransformedTargetForecaster
from sktime.transformations.series.detrend import Deseasonalizer
from sktime.forecasting.arima import ARIMA

forecaster = TransformedTargetForecaster(
    [
        ("deseasonalize", Deseasonalizer(model="multiplicative", sp=12)),
        ("model", ARIMA()),
    ]
)
pipe = ForecastingPipeline(
    [
        ("impute", Imputer()),
        ("log", LogTransformer()),
        ("forecast", forecaster)
    ]
)

#### Step 3: Train and Predict ----
pipe.fit(y, X=None, fh=np.arange(1,13))
pipe.predict(X=None)



"""
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_9484/2011357560.py in <module>
     28 
     29 #### Train and Predict
---> 30 pipe.fit(y, X=None, fh=np.arange(1,13))
     31 pipe.predict(X=None)

~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\sktime\forecasting\base\_base.py in fit(self, y, X, fh)
    178         #####################################################
    179 
--> 180         self._fit(y=y_inner, X=X_inner, fh=fh)
    181 
    182         # this should happen last

~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\sktime\forecasting\compose\_pipeline.py in _fit(self, y, X, fh)
    207         name, forecaster = self.steps[-1]
    208         f = clone(forecaster)
--> 209         f.fit(y, X, fh)
    210         self.steps_[-1] = (name, f)
    211 


~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\sktime\forecasting\base\_base.py in fit(self, y, X, fh)
    178         #####################################################
    179 
--> 180         self._fit(y=y_inner, X=X_inner, fh=fh)
    181 
    182         # this should happen last

~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\sktime\forecasting\compose\_pipeline.py in _fit(self, y, X, fh)
    347         for step_idx, name, transformer in self._iter_transformers():
    348             t = clone(transformer)
--> 349             y = t.fit_transform(y, X)
    350             self.steps_[step_idx] = (name, t)
    351 

~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\sktime\transformations\base.py in fit_transform(self, Z, X)
     86         if X is None:
     87             # Fit method of arity 1 (unsupervised transformation)
---> 88             return self.fit(Z).transform(Z)
     89         else:
     90             # Fit method of arity 2 (supervised transformation)

~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\sktime\transformations\series\detrend\_deseasonalize.py in fit(self, Z, X)
    119             filt=None,
    120             two_sided=True,
--> 121             extrapolate_trend=0,
    122         ).seasonal.iloc[:sp]
    123 

~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\pandas\util\_decorators.py in wrapper(*args, **kwargs)
    205                 else:
    206                     kwargs[new_arg_name] = new_arg_value
--> 207             return func(*args, **kwargs)
    208 
    209         return cast(F, wrapper)

~\.conda\envs\pycaret_dev_sktime_0p7\lib\site-packages\statsmodels\tsa\seasonal.py in seasonal_decompose(x, model, filt, period, two_sided, extrapolate_trend)
    130 
    131     if not np.all(np.isfinite(x)):
--> 132         raise ValueError("This function does not handle missing values")
    133     if model.startswith('m'):
    134         if np.any(x <= 0):

ValueError: This function does not handle missing values
"""