Skip to content

fix(time_series): fix 1-based indexing in make_leads - #493

Open
reginaldalfret wants to merge 1 commit into
Kaggle:masterfrom
reginaldalfret:fix/make-leads-range-offset
Open

reginaldalfret wants to merge 1 commit into
Kaggle:masterfrom
reginaldalfret:fix/make-leads-range-offset

Conversation

@reginaldalfret

Copy link
Copy Markdown

Summary

This PR addresses #467 by fixing the range indexing in make_leads.

Problem

In learntools/time_series/utils.py, make_leads used reversed(range(leads)). For leads=1, range(1) evaluated to 0, generating a single column named {name}_lead_0 containing ts.shift(0) (the contemporaneous unshifted series). For leads=2, it produced lead_1 and lead_0, missing lead_2.

In contrast, make_lags starts at lead_time=1 and generates lag_1 (ts.shift(1)), lag_2, etc.

Solution

  • Updated make_leads to use reversed(range(1, leads + 1)).
  • When leads=1, it now correctly produces {name}_lead_1 with ts.shift(-1).
  • When leads=2, it produces {name}_lead_2 and {name}_lead_1 with ts.shift(-2) and ts.shift(-1), matching the 1-based indexing of make_lags.

Closes #467

Use range(1, leads + 1) instead of range(leads) so that leads start at
lead_1 (shift(-1)) rather than lead_0 (shift(0), which is the unshifted
contemporaneous series). This makes make_leads symmetric with make_lags.

Closes Kaggle#467
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.

Bug in make_leads function

1 participant