[ENH] basic setar-tree module and tests #2890

TinaJin0228 · 2025-06-09T09:54:57Z

Reference Issues/PRs

#2816.

What does this implement/fix? Explain your changes.

In this commit, the major functions for the setar-tree algorithm is implemented, and the basic test file to test them.

Does your contribution introduce a new dependency? If yes, which one?

No.

Any other comments?

This is the initial commit for #2816.

Next steps:

Elaborate on the SETAR-tree algorithm, including an accelerated grid-search for finding the optimal split.
Add tests for corner cases and real-world datasets.
Write documentation and update the API reference page.
Explore implementing separate TAR and SETAR forecasters.

PR checklist

For all contributions

I've added myself to the list of contributors. Alternatively, you can use the @all-contributors bot to do this for you after the PR has been merged.
The PR title starts with either [ENH], [MNT], [DOC], [BUG], [REF], [DEP] or [GOV] indicating whether the PR topic is related to enhancement, maintenance, documentation, bugs, refactoring, deprecation or governance.

For new estimators and functions

I've added the estimator/function to the online API documentation.
(OPTIONAL) I've added myself as a __maintainer__ at the top of relevant files and want to be contacted regarding its maintenance. Unmaintained files may be removed. This is for the full file, and you should not add yourself if you are just making minor changes or do not want to help maintain its contents.

For developers with write access

(OPTIONAL) I've updated aeon's CODEOWNERS to receive notifications about future changes to these files.

aeon-actions-bot · 2025-06-09T09:55:25Z

Thank you for contributing to `aeon`

I have added the following labels to this PR based on the title: [ enhancement ].
I have added the following labels to this PR based on the changes made: [ forecasting ]. Feel free to change these if they do not properly represent the PR.

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 · 2025-06-15T03:17:16Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

MatthewMiddlehurst

Some comments, more to come.

MatthewMiddlehurst · 2025-07-06T20:48:51Z

aeon/forecasting/_setar.py

+    Parameters
+    ----------
+    lag : int, default=10
+        The maximum number of past lags to consider for both the AR models
+        and as the thresholding variable.


You should document the horizon here also, you can reuse the text from other classes..

MatthewMiddlehurst · 2025-07-06T20:49:04Z

aeon/forecasting/_setar.py

+    lag : int, default=10
+        The maximum number of past lags to consider for both the AR models
+        and as the thresholding variable.
+    """


I would add an example usage of the class to the docstring here.

MatthewMiddlehurst · 2025-07-06T20:55:36Z

aeon/forecasting/_setar.py

+
+        for _lag in range(self.lag, 0, -1):
+            if len(y) <= _lag:


just lag should be fine for this loop, no _ needed as its not an attribute.

MatthewMiddlehurst · 2025-07-06T20:56:25Z

aeon/forecasting/_setar.py

+
+class SetarForecaster(BaseForecaster):
+    """


Could you please add a test file for this class and add some functions. If you could generate some expected results from another implementation to ensure correctness that would be good as well.

MatthewMiddlehurst · 2025-07-06T20:57:30Z

aeon/forecasting/_setar.py

+
+class SetarForecaster(BaseForecaster):
+    """


IMO the class name should just be SETAR. I do not think this is used anywhere other than forecasting, and we can easily change if it is.

MatthewMiddlehurst · 2025-07-06T20:57:53Z

aeon/forecasting/__init__.py

Add the regular SETAR to the init as well.

MatthewMiddlehurst

I think this is a good start. Could you post here on what your current plans are for testing correctness i.e. data used and results/implementation to compare against.

You mentioned previously having some issues with implementing global methods using the current framework. Could you also post that here?

Please add the classes to the API documentation in docs/

MatthewMiddlehurst · 2025-07-06T22:14:59Z

aeon/forecasting/_setartree.py

+
+class SetartreeForecaster(BaseForecaster):
+    """


I think SETARTree is Bette as a name. Same reason as SETAR.

I will also modify SetarforestForecaster to SETARForest

MatthewMiddlehurst

I am wondering where we can improve efficiency using numba. Do you know where most time is spent processing in the current implementation?

TinaJin0228 · 2025-07-07T09:57:05Z

Current plan for testing:
Several experiment results are in the blog (https://medium.com/@jintina48/gsoc-experiment-record-f0aa3bd82c18).
The function of fitting and forecasting is ok, but there is still a significant gap of the results, so my current plan is:

Carefully compare my implementation to that of the R codebase, especially the splitting function and the error calculator (which, in my opinion, is the most likely one to have flaws)(though I need more time to locate them)
Insist on first testing on the Chaotic dataset, since it is the simplest and uniform dataset in the paper's evaluation benchmark.

Issues implementing global methods using the current framework:
The current Aeon framework treats multiple time series input as multivariate time series.
Current temporary solution is to pass one time series to the input "x" and others to the exogenous variable, which is a quick fix.

As for the efficiency:
I think the primary computational bottleneck is the find_optimal_split function, which is the major step in building the tree.

TonyBagnall · 2025-07-07T10:25:09Z

aeon/forecasting/_setartree.py

+    def __init__(
+        self,
+        lag: int = 10,
+        horizon: int = 1,


for now, set horizon tag to false and dont p[ass the horizon

TonyBagnall · 2025-07-07T10:25:49Z

aeon/forecasting/_setartree.py

+
+        return self
+
+    def _predict(self, y=None, exog=None):


dont need to check is fitted here, its done in the base class

TonyBagnall · 2025-07-07T10:26:58Z

aeon/forecasting/_setartree.py

+
+        predictions = []
+
+        for _ in range(self.horizon):


dont heed a horizon here. predict should simply predict one ahead based on y. Further prediction horizon predictions are made with the iterative_forecast()

TinaJin0228 · 2025-07-14T09:40:50Z

Progress: I’ve successfully reproduced the results from the paper (with a tiny gap) on the Chaotic dataset, using an independent Python implementation. I’ll update the Aeon branch once I’ve confirmed that the method works within the Aeon framework. Together be done alongside the modification requests mentioned above.

basic setar-tree module and tests

0b987a3

aeon-actions-bot bot added enhancement New feature, improvement request or other non-bug code enhancement forecasting Forecasting package labels Jun 9, 2025

MatthewMiddlehurst marked this pull request as draft June 9, 2025 10:05

TonyBagnall requested a review from MatthewMiddlehurst June 9, 2025 10:11

TinaJin0228 added 2 commits June 14, 2025 22:45

Merge branch 'main' into setar: update

6ae2f44

add setar.ipynb as a temp document of setar forecaster

636713c

TinaJin0228 and others added 6 commits June 15, 2025 17:50

separate SETAR forecaster (demo)

711e63b

Merge branch 'main' into setar

2243b2f

small modifications to pass the checks

3e04609

to pass checks

4673fbf

delete multivariate tag since setar-tree is a univariate forecaster

0ad7059

Merge branch 'main' into setar

3f6db17

MatthewMiddlehurst requested changes Jul 6, 2025

View reviewed changes

MatthewMiddlehurst reviewed Jul 6, 2025

View reviewed changes

TonyBagnall requested changes Jul 7, 2025

View reviewed changes

[ENH] basic setar-tree module and tests #2890

Are you sure you want to change the base?

[ENH] basic setar-tree module and tests #2890

Uh oh!

Conversation

TinaJin0228 commented Jun 9, 2025

Reference Issues/PRs

What does this implement/fix? Explain your changes.

Does your contribution introduce a new dependency? If yes, which one?

Any other comments?

PR checklist

For all contributions

For new estimators and functions

For developers with write access

Uh oh!

aeon-actions-bot bot commented Jun 9, 2025

Thank you for contributing to aeon

Uh oh!

review-notebook-app bot commented Jun 15, 2025

Uh oh!

MatthewMiddlehurst left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MatthewMiddlehurst left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MatthewMiddlehurst left a comment

Choose a reason for hiding this comment

Uh oh!

TinaJin0228 commented Jul 7, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TinaJin0228 commented Jul 14, 2025

Uh oh!

Uh oh!

Thank you for contributing to `aeon`