README.md

---
license: cc-by-4.0
task_categories:
- time-series-forecasting
tags:
- time-series
- forecasting
- anomaly-detection
- classification
- TSLib
- tsfile
- modality:timeseries
- timeseries
- format:tsfile
pretty_name: MSL (TsFile)
size_categories:
- 1M<n<10M
configs:
- config_name: default
  data_files:
  - split: train
    path: MSL_train.tsfile
  - split: test
    path: MSL_test.tsfile
  - split: train_labels
    path: labels/MSL_train_label.csv
  - split: test_labels
    path: labels/MSL_test_label.csv
modality:
- timeseries
- tabular
- text
language:
- en
---

# MSL (TsFile)

Apache TsFile version of the `MSL` anomaly-detection subset of
[`thuml/Time-Series-Library`](https://huggingface.co/datasets/thuml/Time-Series-Library).

## Overview

Mars Science Laboratory rover telemetry (55 channels) for anomaly detection.

- **Train:** 58,317 rows.
- **Test:** 73,729 rows.
- **Channels:** 55.

The train and test segments are stored as two separate TsFiles
(`MSL_train.tsfile` / `MSL_test.tsfile`), preserving the original split. Labels
are stored as separate CSV sidecar files in `labels/`.

## Schema (TsFile structure)

- **Time** (INT64, milliseconds) - row index * 1000 ms. The `.npy` source has no
  timestamp, so this is a monotone 1 Hz proxy axis.
- **FIELD** (55 channels, FLOAT) - the sensor/metric channels.
- **Labels** (CSV sidecar) - `labels/MSL_train_label.csv` and
  `labels/MSL_test_label.csv` contain `Time,label`. The train label file is all
  0; the test label file carries the per-timestep anomaly labels.

No channels or rows are dropped. Labels are not stored inside the TsFile feature
tables.

## Usage

Read the `.tsfile` files with the Apache TsFile Java or Python SDK. Join labels
from the sidecar CSV files by `Time` within the matching split.

## Source & license

- Original dataset: https://huggingface.co/datasets/thuml/Time-Series-Library (subset `MSL`)
- Author / publisher: thuml (Tsinghua University)
- Paper: https://arxiv.org/abs/2407.13278
- License: CC BY 4.0