Working with TimedeltaIndex [newbie]

[mantielero]

I am new to sktime (and in fact I have spent a significant time without doing any python).

I have started working with forecasting. I am using a dataset that, when imported, it looks like:

>>> print(df.iloc[:,0].index)
0 days 00:00:00          -0.254
0 days 00:00:00.002000   -0.254
0 days 00:00:00.004000   -0.254
0 days 00:00:00.006000   -0.254
0 days 00:00:00.008000   -0.264
                          ...  
0 days 00:00:09.990000   -0.044
0 days 00:00:09.992000   -0.034
0 days 00:00:09.994000   -0.034
0 days 00:00:09.996000    0.024
0 days 00:00:09.998000    0.005
Freq: 2L, Name: I, Length: 5000, dtype: float64

The index column type seems to be:

<class 'pandas.core.indexes.timedeltas.TimedeltaIndex'>

I am able to fit it like:

forecaster = NaiveForecaster(strategy="last", sp=12)
forecaster.fit(df.iloc[:,0])

But when I try to predict:

fh = pd.timedelta_range(start='0 days 00:00:10.000000', 
                        end ='0 days 00:00:14.000000', 
                        freq="2L")
y_pred = forecaster.predict(fh)

I get the error:

  File "/home/jose/.local/lib/python3.10/site-packages/sktime/utils/datetime.py", line 220, in _shift
    assert isinstance(x, (pd.Period, pd.Timestamp, int, np.integer)), type(x)
AssertionError: <class 'pandas._libs.tslibs.timedeltas.Timedelta'>

Why is that? How should I do it?

[fkiraly]

hm, there could be a bug (or a mistake)?

What would help a lot is full code, optimally using only sktime and dummy data, that breaks in the same way.

We can then more easily investigate what is going on here and help you, much appreciated.