better formulation to wrap-up

[psteinb]

The last sentence of the cross_validation_time.ipynb notebook felt a bit convoluted. I tried my best to improve the formulation.

[ArturoAmorQ]

Thanks for the PR @psteinb! Here are a couple of comments. Also, the linter fail in the CI is because it's configured to 80 characters max.

Being said that, reworking this whole notebook has been in my list of to do's for a long time. In particular, we want to use a more realistic dataset, give interpretation to resulting R2 and MSE scores, and mention the actual good practices for modeling, e.g. aligning the test size of TimeSeriesSplit with the forecasting task. In case you want to contribute, feel free to submit a more in depth PR.

[ArturoAmorQ]

I don't think "out of the shelves" is grammatically correct nor a common term for the students to know. Maybe we can rephrase the whole paragraph.

[ArturoAmorQ]

Hi @psteinb are you still working on this PR? Would you mind adding a small mention in the conclusions in the line of:

However, this does not mean that scikit-learn is useless for time-series tasks. In fact, scikit-learn offers useful utilities for time-series analysis (interested readers can see for instance the Time-related feature engineering example in the documentation), and its models can yield even better results when combined with other specialized libraries.

[ArturoAmorQ]

Thanks for your contribution @psteinb!
My comment here can wait for a further PR.