[DOC] Rework getting started guide and single problem forecasting loaders

[TonyBagnall]

fixes #2246 part of #1518

This has expanded to tidying up the single problem data loaders for forecasting. So its in two related parts

datasets._single_problem_loaders

There are seven baked in forecasting data sets and were eight loaders.

1. I have removed load_macroeconomic because it was just a wrapper for the statsmodels loader
2. For five univariate series loaders, I have added a return_array boolean argument that defaults to true. This makes it load data as an np.ndarray, if false it returns a pd.Series
3. There are two multivariate loaders, uschange and longley. These adopted a structure like this

y, X = load_longley(y_series = "Consumption")

firstly returning y, X is opposite to collections, and secondly there seems no need to split the data in the loader, the user can surely do that themselves. Changed to

data =load_longley()

returns numpy array with axis == 0, i.e. n_channels, n_timepoints

and

data =load_longley(return_array = False)

returns a data frame with axis == 1 and all the column names set as before.

Read me

Split along series/collection estimators and adding an example for each module, including experimental. It makes it longer, maybe we dont want it, but its a good top level intro imo, will link for further details

First version done, highlighted there is no anomaly detection notebook, see #1960 and the transformers notebooks need an overhaul, but thats future work. The main goal is to get things ready for the new forecasting base class

[aeon-actions-bot]

Thank you for contributing to aeon

I did not find any labels to add that did not already exist. If the content of your PR changes, make sure to update the labels accordingly.

The Checks tab will show the status of our automated tests. You can click on individual test runs in the tab or "Details" in the panel below to see more information if there is a failure.

If our pre-commit code quality check fails, any trivial fixes will automatically be pushed to your PR unless it is a draft.

Don't hesitate to ask questions on the aeon Slack channel if you have any.

PR CI actions

These checkboxes will add labels to enable/disable CI functionality for this PR. This may not take effect immediately, and a new commit may be required to run the new configuration.

• Run pre-commit checks for all files
• Run mypy typecheck tests
• Run all pytest tests and configurations
• Run all notebook example tests
• Run numba-disabled codecov tests
• Stop automatic pre-commit fixes (always disabled for drafts)
• Disable numba cache loading
• Push an empty commit to re-run CI checks

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[MatthewMiddlehurst]

we are defaulting, but it is not universal, see anomaly detection.

Where in anomaly detection? I am referring just to inputs, dont really care what format they want the data internally

You cant load into a dataframe with n_channels, n_timepoints and keep the column names. I have left the pandas stuff for legacy reasons really. I think I would rather remove these loaders completely than process to have series in rows in a dataframe.

cant you just transpose the dataframe? seems very odd if that removes the indicies

[TonyBagnall]

we are defaulting, but it is not universal, see anomaly detection.

Where in anomaly detection? I am referring just to inputs, dont really care what format they want the data internally

You cant load into a dataframe with n_channels, n_timepoints and keep the column names. I have left the pandas stuff for legacy reasons really. I think I would rather remove these loaders completely than process to have series in rows in a dataframe.

cant you just transpose the dataframe? seems very odd if that removes the indicies

I guess I can, seems odd to do so and I thought the examples made it clear, but sure, Need to rewrite the tests that use col_names

[MatthewMiddlehurst]

I remember some previous changes where we tried to remove n_timepoints, n_channels as much as we could. Think it was df-list and the other dataframe one for collections. May have been some other changes removing references to it as well.

Seems better to completely follow one format and leave changing that to the axis stuff if thats how we want it

[TonyBagnall]

we are defaulting, but it is not universal, see anomaly detection.

Where in anomaly detection? I am referring just to inputs, dont really care what format they want the data internally

You cant load into a dataframe with n_channels, n_timepoints and keep the column names. I have left the pandas stuff for legacy reasons really. I think I would rather remove these loaders completely than process to have series in rows in a dataframe.

cant you just transpose the dataframe? seems very odd if that removes the indicies

yes but you cannot then extract series X["channel_name"], anyway, I have done it and removed the test for column names

[MatthewMiddlehurst]

You would just use X.loc["channel_name"] im pretty sure

https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.loc.html

[TonyBagnall]

You would just use X.loc["channel_name"] im pretty sure

https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.loc.html

It was more I didn't want to rewrite the notebooks that used plot_series. But I've ditched that now

[hadifawaz1999]

just small questions, if you want to keep them for later i dont mind

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

hadifawaz1999 commented on 2024-11-03T12:51:53Z
----------------------------------------------------------------

i would think its better to add a section per task to use the load any dataset function, load_classification load_regression etc. what do u think ?

TonyBagnall commented on 2024-11-03T16:46:53Z
----------------------------------------------------------------

yes I agree, but maybe not in this PR? Really only wanted to do the getting_started.md, then will work through notebooks module by module, starting with datasets

[TonyBagnall]

yes I agree, but maybe not in this PR? Really only wanted to do the getting_started.md, then will work through notebooks module by module, starting with datasets

---

View entire conversation on ReviewNB

[hadifawaz1999]

lgtm