I'm exploring hierarchal forecasting for time series in python and came across scikit-hts library. I installed the package using pip install scikit-hts. I followed the steps mentioned in this tutorial. I created the hierarchal tree and came to the modelling part. I want to predict bottom level forecast using Facebook prophet algorithm. I ran the following code
clf = hts.HTSRegressor(model='prophet', revision_method='AHP', n_jobs=0)
model = clf.fit(hierarchy_train, hierarchy)
This has thrown the warning message,
Fitting models: 0%| | 0/46 [00:00<?, ?it/s]prophet model requires fbprophet to
work. Exiting.Install it with: pip install scikit-hts[prophet]
Fitting models: 0%| | 0/46 [00:00<?, ?it/s]
and this error message,
AttributeError: 'NoneType' object has no attribute 'fit'
I tried to install every additional package from here. But it is throwing the following error message for every package (here only shown for arima model)
zsh: no matches found: scikit-hts[arima]
Can someone please explain why it is happening? Is it because that the packages are discontinued or is there a issue in my installation process?
Edit:
I have updated my version of scikit-hts but still
pip install scikit-hts[x](wherex=all/geo/arima/prophet) didn't workI ran my code in the jupyter notebook. All models worked except for prophet. Here is the code.
clf = hts.HTSRegressor(model='x', revision_method='OLS', n_jobs=20) model = clf.fit(hierarchy_train, hierarchy)
This code ran for x = auto_arima, holt_winters, sarimax but not for prophet
I got the following error when I tried prophet
AttributeError: 'NoneType' object has no attribute 'fit'
Why is the model not running for prophet?
- I don't know if it is relevant or not but I have installed statsmodels packages for arima and other smoothening models. I have installed fb-prophet on my system but I didn't import in my notebook. I have only done
import hts