Skip to content

fix: preserve the full MFLES seasonal cycle - #1215

Open
mrlonely (mameikagou) wants to merge 1 commit into
Nixtla:mainfrom
mameikagou:codex/fix-mfles-multiseasonal-cycle
Open

mrlonely (mameikagou) wants to merge 1 commit into
Nixtla:mainfrom
mameikagou:codex/fix-mfles-multiseasonal-cycle

Conversation

@mameikagou

Copy link
Copy Markdown

Summary

MFLES now stores the combined seasonal pattern over the least common multiple of all configured periods. This prevents forecasts from truncating the pattern when the largest period is not divisible by the smaller periods.

Closes #938.

Tests

  • Added a regression test for seasonal periods 2 and 5
  • Verified the regression fails before the fix and passes afterward
  • Verified single, duplicate, and nested seasonal periods, plus serialization
  • Ruff and git diff --check

The seasonal buffer is now proportional to the periods’ LCM; configurations where one period divides another retain their existing size.

@CLAassistant

CLAassistant commented Aug 21, 2026 •

Copy link
Copy Markdown

CLA assistant check
All committers have signed the CLA.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[MFLES] Seasonality prediction incorrect when forecasting multi-seasonal time series beyond one period

2 participants