README.md

metadata

license: apache-2.0
tags:
  - forecasting
  - benchmark
  - gifteval
  - tsfile
  - format:tsfile
  - timeseries
task_categories:
  - time-series-forecasting
pretty_name: m4_weekly (GIFT-Eval, TsFile)
size_categories:
  - n<1K

m4_weekly (TsFile)

Apache TsFile version of the m4_weekly subset of GIFT-Eval.

Overview

GIFT-Eval is a benchmark for general time-series forecasting, covering 23 datasets (≈144,000 series and 177M data points) across seven domains, ten frequencies, and a range of forecast horizons. This repository contains a single subset of that benchmark.

m4_weekly — Weekly-frequency series from the M4 forecasting competition.

All .tsfile files are stored under data/.

Schema (TsFile structure)

• Time (INT64, milliseconds) — the timestamp of each observation.
• Each series is stored as a TsFile device; per-series identifiers from GIFT-Eval are carried as TAG columns, and the observed values are FIELD columns.

Usage

Read the .tsfile files with the Apache TsFile Java or Python SDK.

Source & license

• Original dataset: Salesforce/GiftEval — subset m4_weekly
• Paper: GIFT-Eval (arXiv:2410.10393)
• Code: https://github.com/SalesforceAIResearch/gift-eval
• License: apache-2.0

If you use this data, please cite GIFT-Eval:

@article{aksu2024giftevalbenchmarkgeneraltime,
  title={GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluation},
  author={Taha Aksu and Gerald Woo and Juncheng Liu and Xu Liu and Chenghao Liu and Silvio Savarese and Caiming Xiong and Doyen Sahoo},
  journal={arXiv preprint arXiv:2410.10393},
  year={2024}
}