Datasets Guide

Overview

In this open-source anomaly detection benchmarking project, we only work with publicly available published dataset:

• MIT/Stanford Data-driven Prediction of Battery Cycle Life Dataset

• Tohoku Dataset

Data management toolstack

DuckDB: The data management is this project is implemented using DuckDB (Link: https://duckdb.org/). To access the dataset curated in this project, you would need to first install duckdb:

uv pip install duckdb

Git Large File Storage (LFS)

Because large files cannot be committed and versioned directly in Git repositories, we used Git Large File Storage (LFS) for versioning large dataset in this project (Link: https://git-lfs.com/). You can follow the documentation there to setup Git LFS for your local development.

Contributions

We welcome contributions to enhance our open-source anomaly detection benchmarking suite. New datasets, use-cases, or extensions to existing repositories are highly encouraged. If you are interested in contributing, feel free to reach out or submit a pull request.

Example Dataset

List of Dataset

• MIT/Stanford Data-driven Prediction of Battery Cycle Life Dataset
  • Battery Chemistry Description
  • Train and Test Dataset
  • Dataset Information
  • Features Description
  • Data Quality Assessment
  • Sample Dataset for Cell 2017-05-12_5_4C-70per_3C_CH17
  • Citation
• Tohoku Dataset
  • Battery Chemistry Description
  • Feature Description