CODEWITHSUNDEEP
pythondatedatetimepredictarima
hellobello
hellobello

Reputation: 1

Unable to predict values for a given input date in ARIMA using predict() function

I am using the following code for arima to predict values of output for the given date value from 2018-10-17 to 2018-10-22. But I get the error as

ERROR: "int() argument must be a string, a bytes-like object or a number, not 'Timestamp'".

model = ARIMA(df, order=(5,1,0))
model = model.fit(disp=0)
pred = model.get_prediction(start=pd.to_datetime('2018-10-17'),
                            end=pd.to_datetime('2018-10-22'),
                            dynamic=True)

Doesn't predict() accept datetime objects? It says it does in the documentation.

Upvotes: 0

Views: 2045

Answers (1)

unutbu
unutbu

Reputation: 880887

Assuming you are using statsmodels.tsa.statespace.sarimax.SARIMAXResults.get_prediction, notice that start and end expect an int, str, or datetime. Per the docs, the str can be a date string, so you could use

pred = model.get_prediction(start='2018-10-17', 
                            end='2018-10-22',
                            dynamic=True)

Pandas Timestamps are not the same as Python datetime.datetime objects.

If you wish to use a datetime, use the Timestamp.to_pydatetime() method to convert Pandas Timestamps to Python datetime.datetime objects:

In [181]: pd.to_datetime('2018-10-17')
Out[181]: Timestamp('2018-10-17 00:00:00')

In [180]: pd.to_datetime('2018-10-17').to_pydatetime()
Out[180]: datetime.datetime(2018, 10, 17, 0, 0)

or simply create the datetime.datetime object directly:

In [184]: import datetime as DT

In [185]: DT.datetime(2018,10,17)
Out[185]: datetime.datetime(2018, 10, 17, 0, 0)

pred = model.get_prediction(start=DT.datetime(2018, 10, 17), 
                            end=DT.datetime(2018, 10, 22),
                            dynamic=True)

Upvotes: 1

Related Questions