I am trying to fit a time series to some data I have, and would like to forecast into the future. I have 2 questions: 1) is the model I use for this particular time series appropriate? and 2) Why, when I try to forecast, am I getting an error message?
The data more or less looks linear with some slight seasonality, so I thought statsmodels UnobservedCompoents model would give a good fit. The model appears to give a good fit as shown in the figures below. I have included the code I used to make these figures.
import matplotlib.pyplot as plt
import matplotlib
matplotlib.style.use('ggplot')
import numpy as np
import pandas as pd
from statsmodels.tsa.statespace.structural import UnobservedComponents
endog = pd.read_csv('file_name', parse_dates=True, index_col=0)
model=UnobservedComponents(endog, 'local level').fit()
fig = model.plot_components(figsize=(10, 10))
The dataframe, endog, looks like this:
The model appears to fit the data quite well, but I am getting an error message when I try to forecast 100 steps into the future with the following command.
forecast=model.forecast(steps=100)
The full error is given below. Can anyone tell me why the .forecast method works with other statsmodels time series models, but is giving me an error here? Thanks for your help.
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-16-c04a026facb5> in <module>()
----> 1 forecast=model.forecast(steps=100)
/Users/deeps/anaconda/lib/python3.5/site-packages/statsmodels-0.8.0-py3.5- macosx-10.5-x86_64.egg/statsmodels/base/wrapper.py in wrapper(self, *args, **kwargs)
93 obj = data.wrap_output(func(results, *args, **kwargs), how[0], how[1:])
94 elif how:
---> 95 obj = data.wrap_output(func(results, *args, **kwargs), how)
96 return obj
97
/Users/deeps/anaconda/lib/python3.5/site-packages/statsmodels-0.8.0-py3.5- macosx-10.5-x86_64.egg/statsmodels/tsa/statespace/structural.py in forecast(self, steps, exog, **kwargs)
1547 """
1548 return super(UnobservedComponentsResults, self).forecast(
-> 1549 steps, exog=exog, **kwargs)
1550
1551 def summary(self, alpha=.05, start=None):
/Users/deeps/anaconda/lib/python3.5/site-packages/statsmodels-0.8.0-py3.5-macosx-10.5-x86_64.egg/statsmodels/tsa/statespace/mlemodel.py in forecast(self, steps, **kwargs)
2373 else:
2374 end = steps
-> 2375 return self.predict(start=self.nobs, end=end, **kwargs)
2376
2377 def simulate(self, nsimulations, measurement_shocks=None,
/Users/deeps/anaconda/lib/python3.5/site-packages/statsmodels-0.8.0-py3.5-macosx-10.5-x86_64.egg/statsmodels/tsa/statespace/structural.py in predict(self, start, end, exog, dynamic, **kwargs)
1476
1477 # Handle end (e.g. date)
-> 1478 _start = self.model._get_predict_start(start)
1479 _end, _out_of_sample = self.model._get_predict_end(end)
1480
/Users/deeps/anaconda/lib/python3.5/site-packages/statsmodels-0.8.0-py3.5-macosx-10.5-x86_64.egg/statsmodels/tsa/base/tsa_model.py in _get_predict_start(self, start)
129 (str(start), str(dtstart)))
130
--> 131 self._set_predict_start_date(start)
132 return start
133
/Users/deeps/anaconda/lib/python3.5/site-packages/statsmodels-0.8.0-py3.5-macosx-10.5-x86_64.egg/statsmodels/tsa/base/tsa_model.py in _set_predict_start_date(self, start)
104 if start == len(dates):
105 self.data.predict_start = datetools._date_from_idx(dates[-1],
--> 106 1, self.data.freq)
107 elif start < len(dates):
108 self.data.predict_start = dates[start]
/Users/deeps/anaconda/lib/python3.5/site-packages/statsmodels-0.8.0-py3.5-macosx-10.5-x86_64.egg/statsmodels/tsa/base/datetools.py in _date_from_idx(d1, idx, freq)
77 offset. For now, this needs to be taken care of before you get here.
78 """
---> 79 return d1 + idx * _freq_to_pandas[freq]
80
81 def _idx_from_dates(d1, d2, freq):
TypeError: unsupported operand type(s) for *: 'int' and 'NoneType'
