Skip to content

feat: v4-M2 — snapshot day tuning + validation fix + release docs#24

Merged
shaypal5 merged 1 commit into
mainfrom
feat/v4-m2-release-docs
Apr 29, 2026
Merged

feat: v4-M2 — snapshot day tuning + validation fix + release docs#24
shaypal5 merged 1 commit into
mainfrom
feat/v4-m2-release-docs

Conversation

@shaypal5

Copy link
Copy Markdown
Contributor

Summary

  • Snapshot day 14 → 10: After the lead-source boost stacking fix in PR feat: Milestone v4-M1 — engine changes and build pipeline #21 changed population latent distributions, the leakage trap boost dropped below the 0.03 threshold at day 14. Lowering to day 10 gives legitimate features less signal (shorter observation window), restoring the detectable boost (0.034).
  • StandardScaler in validation LR: sklearn LogisticRegression was not converging reliably without feature scaling. Added StandardScaler to _fit_lr() in validate_v4_dataset.py.
  • v4 dataset generated and validated: 1000 rows × 18 columns, 30.0% conversion rate, baseline AUC 0.652, leakage trap boost 0.034. All 7 mandatory checks pass. Dataset + RELEASE_v4.md committed to leadforge-datasets-private.
  • .agent-plan.md updated: v4-M1 and v4-M2 marked complete.

Test plan

  • CI passes (existing 609 tests)
  • python scripts/build_v4_snapshot.py /tmp/test_v4.csv produces valid CSV
  • python scripts/validate_v4_dataset.py /tmp/test_v4.csv exits 0

🤖 Generated with Claude Code

- Change SNAPSHOT_DAY from 14 to 10 to ensure leakage trap boost ≥ 0.03
  after the lead-source boost stacking fix changed population distributions
- Add StandardScaler to validate_v4_dataset.py LR pipeline to fix
  convergence warnings
- Update .agent-plan.md to mark v4-M1 and v4-M2 complete

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Copilot AI review requested due to automatic review settings April 29, 2026 18:27
@shaypal5 shaypal5 added type: docs Documentation or narrative changes type: chore Tooling, config, repo hygiene labels Apr 29, 2026
@github-actions

Copy link
Copy Markdown

pr-agent-context report:

No unresolved review comments, failing checks, or actionable patch coverage gaps were found on PR
#24. Treat this PR as all clear unless new signals appear.

Run metadata:

Tool ref: v4
Tool version: 4.0.20
Trigger: pull request opened
Workflow run: 25126607354 attempt 1
Comment timestamp: 2026-04-29T18:27:57.148098+00:00
PR head commit: 103bf93aefa786c77ffece9eb97ba5b2bc3135a9

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR tunes the v4 dataset build/validation utilities to restore the intended “leakage trap” signal after earlier population distribution changes, and updates internal milestone tracking accordingly.

Changes:

  • Adjust the v4 snapshot window from day 14 to day 10 in the build script.
  • Add StandardScaler to the validation script’s LogisticRegression fitting to improve convergence stability.
  • Update .agent-plan.md milestone status and recorded v4 validation metrics.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File Description
scripts/validate_v4_dataset.py Scales numeric features before fitting LR to stabilize convergence for AUC/leakage-trap checks.
scripts/build_v4_snapshot.py Switches snapshot window constant/docs from day-14 to day-10.
.agent-plan.md Marks v4 milestones complete and updates reported metrics/notes.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@shaypal5 shaypal5 merged commit 65fb508 into main Apr 29, 2026
9 checks passed
@shaypal5 shaypal5 deleted the feat/v4-m2-release-docs branch April 29, 2026 18:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

type: chore Tooling, config, repo hygiene type: docs Documentation or narrative changes

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants