sample/synthetic_demand.csv is deterministic synthetic data generated by
scripts/generate_sample_data.py. It is not derived from the Walmart China dataset and is
provided only to exercise the software and tests.
precomputed/synthetic-smoke/ is a lightweight output snapshot generated from that same public
synthetic data by the green MATLAB R2024b workflow. Its provenance is recorded alongside the
files. It is integration evidence, not a private-data or publication-grade benchmark.
The original thesis dataset is not distributed because its publication rights have not been
confirmed. It is not covered by the repository's MIT license. Store authorized private data
outside the repository and set ADAPTFORECAST_DATA_DIR to that directory.
Canonical CSV columns:
| Column | Meaning |
|---|---|
date |
ISO date for the demand target |
category |
Product category identifier |
sales |
Observed target-day unit demand |
avg_temp, max_temp, min_temp |
Target-day weather temperatures |
humidity |
Target-day relative humidity in percent |
rainfall |
Target-day precipitation |
The forecasting contract uses past sales and weather known or forecast for the target date. It does not use future observed sales.