time-series-analysis

The Aggregator incorrectly assumes that all hierarchy levels are named, in contrast to the specification of pd_multiindex_hier which does not. The estimator breaks when receiving input where at least one level is not named.

#2432 proposes a temp workaround, but this should be:

• cleaned up so the inner logic of Aggregator no longer assumes it
• tested, so we ensure in all estimators t

The stumpy.snippets feature is now completed in #283 which follows this work:

• The original published paper
• The supporting website with data examples

We have a rough notebook t

On MacOS, the tslearn.datasets does not work out-of-the-box.

In order to make it work, you need to apply the following steps:

1. Go to your finder
2. run "/Apps/Python/Install Certificates.command". This basically installs the certifi package with pip.

Perhaps we should add this to the documentation page of our datasets module?

We have a lot of antiquated docstrings that don't render well into ReadTheDocs. A kind of grunge (but incredibly useful) task would be to refactor these docstrings into proper ReadTheDocs format. This would allow us to render them effectively...

Steps:

1. Create notebook example hello_scikit-hts of using scikit-hts package (similar to hello_divinity)
2. Make a pull request at timeseries-notebooks

In the readme maybe separate automatic forecast into auto_ets and auto_arima, adding an example for the last:

model = auto_arima(log_air_passengers; seasonal = 12)

And in documentation improve the automatic forecast section adding working examples of auto-arima. Saying that is possible to fix an integration order, play with the parameters...