FlexMeasures / flexmeasures / 25572223980

80%

push

github

Fix forecast regressor filtering by belief time (#2134)

* fix(forecasting): apply issue-time aware regressor selection

Split regressor loading around forecast issue time so training and inference keep available beliefs around the boundary while preserving deterministic source selection. Add deterministic integration and unit coverage to guard this behavior.

Made-with: Cursor
Signed-off-by: Ragnar-the-mighty <163057491+Ragnar-the-mighty@users.noreply.github.com>

* docs(changelog): add entry for issue-time regressor filtering fix

Document the forecasting bugfix for issue-time-aware regressor selection and reference the PR from the changelog.

Made-with: Cursor
Signed-off-by: Ragnar-the-mighty <163057491+Ragnar-the-mighty@users.noreply.github.com>

* docs(changelog): describe regressor fix by belief time

Context:
- PR #2134 should use FlexMeasures terminology and point to the correct PR.
Change:
- Move the regressor filtering entry to the current v0.33.0 bugfix section.
- Replace issue-time wording with forecast belief_time wording.

Co-authored-by: Cursor <cursoragent@cursor.com>
Signed-off-by: Ragnar-the-mighty <163057491+Ragnar-the-mighty@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* forecasting: clarify regressor filtering tests

Context:
- PR review asked for clearer test data intent and reuse of existing index helpers.
Change:
- Annotate synthetic belief rows and monkeypatch capture intent.
- Replace local forecast alignment helper with simplify_index.

Signed-off-by: Ragnar-the-mighty <163057491+Ragnar-the-mighty@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* forecasting: document regressor horizon test intent

Context:
- PR review asked why the integration forecasts should differ and how the regressor horizons are set up.
Change:
- Clarify the deterministic accuracy expectation in the docstring.
- Document the 65-minute regressor belief horizon versus the 60-minute forecast horizo... (continued)

20 of 22 new or added lines in 1 file covered. (90.91%)

12382 of 15526 relevant lines covered (79.75%)

0.8 hits per line