How can I get the current version working within Anaconda Jupyter?

[chivalry]

Question Do you have any advice for installing pmdarima under Anaconda and Jupyter? I'm somewhat new to those environments and the one solid answer installed an older version. I then tried setting up my own Anaconda package, but that's erroring out when I try to install it. I finally tried installing from your GitHub repository, but that's also erroring out.

Versions (if necessary) From within my Jupyter notebook:

Darwin-19.0.0-x86_64-i386-64bit Python 3.7.4 (default, Aug 13 2019, 15:17:50) [Clang 4.0.1 (tags/RELEASE_401/final)] pmdarima 1.2.0 NumPy 1.17.2 SciPy 1.3.1 Scikit-Learn 0.21.3 Statsmodels 0.10.1

[tgsmith61591]

This is kind of a multi-part answer, but there's an easy solution. First of all, in #174 we're working on getting a Conda package deployed when we release. We have no affiliation with Saravji so I think he's just repackaged our code to host on Conda, which is fine, but take care when downloading it, since I can't vouch for what's in the tarball.

Getting it working in conda (which is what most of us use when developing) is pretty straight forward:

$ conda create -n pmdarima-example -q python=3.7
$ conda activate pmdarima-example
$ pip install pmdarima

To allow jupyter to find and select your env, you'll need nb_conda, as suggested in this answer on SO:

$ conda install nb_conda jupyter

Then when you start your notebook, your environment should appear in the Kernel > Change Kernel tab:

Select it, then try importing:

[chivalry]

Thanks so much. That worked great. I really appreciate the help. Now I can follow through on your Use cases in the documentation.

BTW, I also really appreciate the Tips to using auto_arima docs. I'm a programmer first, but have been asked to analyse sales for projections, and that one page made so many things click that three or four classes on time series forecasting hadn't.

[tgsmith61591]

I'm glad to hear it! Thanks for finding and fixing issues as you see them.