README.md

metadata

license: cc-by-4.0
annotations_creators:
  - no-annotation
language_creators:
  - found
multilinguality:
  - monolingual
source_datasets:
  - original
task_categories:
  - time-series-forecasting
task_ids:
  - univariate-time-series-forecasting
tags:
  - tsfile
  - format:tsfile
  - timeseries
  - time-series
pretty_name: sunspot (TsFile format)
configs:
  - config_name: default
    data_files:
      - split: train
        path: '*.tsfile'

sunspot (TsFile format)

A single very long daily time series of sunspot numbers from 1818-01-08 to 2020-05-31.

This repository contains the full source .tsf series from the Monash Time Series Forecasting Repository converted to Apache TsFile format.

Summary

• Source dataset: Monash-University/monash_tsf
• Original source: https://zenodo.org/record/4654773
• Monash subset: sunspot
• Modalities: Time-series
• Source series: 1
• Rows: 73,924 flattened timestamped observations
• Frequency: daily
• Forecast horizon metadata: not specified
• Missing-values metadata: True
• Equal-length metadata: True
• Missing target values preserved as NaN: 3,240
• Series length range: 73,924 to 73,924
• TsFile output: 1 file (sunspot.tsfile)

Files

• sunspot.tsfile

TsFile Schema

Column	Role	TsFile type
Time	TIME	INT64
series_id	TAG	STRING
series_name	TAG	STRING
start_timestamp	TAG	STRING
target	FIELD	FLOAT

Conversion Notes

• Each source .tsf data row is stored as one TsFile device.
• Source .tsf attributes are stored as TAG columns.
• The target series values are flattened into timestamped rows and stored as a FLOAT FIELD.
• Time is synthesized from the source start timestamp and the .tsf frequency metadata, with millisecond precision.
• Large outputs may be sharded by the TsFile conversion tool; all listed shards belong to the same logical table sunspot.

Usage

Install the Apache TsFile Python SDK (pip install tsfile) and read a converted file:

from pathlib import Path
from tsfile import TsFileReader

path = Path("sunspot.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    print("columns:", [c.get_column_name() for c in table.get_columns()])
    with reader.query_table(table_name, ["target"], batch_size=1024) as result:
        batch = result.read_arrow_batch()
        if batch is not None:
            print(batch.to_pandas().head())