Reproducible guides and independent study materials

A DuckDB workflow for pinning source data, testing its contract, defining a prediction moment and exporting chronological model features.

A pandas and scikit-learn workflow for validation-led model selection, leakage-safe preprocessing, untouched holdout evaluation and reviewable artefacts.

An independent base-R audit that recalculates holdout errors, diagnoses performance by time group and block-bootstraps the paired model improvement.

Turn reviewed Python predictions and an independent R audit into a validated JSON contract and an accessible, progressively enhanced diagnostic report.

Turn a pinned analytical dataset and prediction-time contract into a typed, fail-closed quality gate that emits a deterministic machine-readable report.

Move a measured, stable numerical kernel behind a validated NumPy boundary, then prove the native result against a readable Python reference before trusting it.

Build and audit a reproducible JuMP mixed-integer optimisation that turns fixed model predictions into a synthetic, constraint-aware monitoring plan.