---
license: mit
task_categories:
  - time-series-forecasting
  - other
tags:
  - retail
  - inventory-optimization
  - predict-then-optimize
  - chronos-2
  - operations-research
  - stochastic-programming
pretty_name: Retail Forecast-Optimize Benchmark
size_categories:
  - 1K<n<10K
---

# Retail Forecast-Optimize Benchmark

Benchmark dataset for **predict-then-optimize** retail inventory decisions.

## Contents

| File pattern | Description |
|--------------|-------------|
| `{scenario}_s{seed}.json` | Full policy comparison per scenario instance |
| `samples/sample_*.json` | Retail scenario definitions |
| `eval_results.json` | Aggregated leaderboard metrics |
| `chronos-2-live-forecasts/` | Live Chronos-2 outputs from HF Jobs (when available) |

## Scenarios

- Baseline Operations
- Promotion Spike
- Holiday Surge
- Oil Price Shock
- Supply Disruption

## Source Data

- [t4tiana/store-sales-time-series-forecasting](https://huggingface.co/datasets/t4tiana/store-sales-time-series-forecasting)
- [Jacoblian/RetailOpt-190](https://huggingface.co/datasets/Jacoblian/RetailOpt-190)

## Forecast Model

[amazon/chronos-2](https://huggingface.co/amazon/chronos-2) — 120M parameter time-series foundation model with quantile forecasting.

## Primary Metrics

Stockout cost, holding cost, waste, service level, total inventory cost, decision regret.

## License

MIT