data-cleaning-annotation-workflow_skill

Overview

This skill provides a complete, repeatable workflow for preparing time series datasets (Energy, Manufacturing, Climate) from Kaggle to a Data Annotation platform. It covers discovery, download, pandas-based cleaning, RAW upload with metadata, column configuration (Time/Target/Covariate/Group), unit assignment, and bulk group tagging. The goal is to move datasets to CLEAN status reliably and reproducibly.

How this skill works

You locate a suitable Kaggle dataset, download CSVs (browser or Kaggle CLI), and run a cleaning script that handles whitespace, duplicates, missing values, and timestamp conversion. Upload the original CSV as RAW with descriptive metadata, then upload the cleaned CSV and configure column types and units. Finally, select all variables and apply all group tags in bulk so the dataset processes to CLEAN status on the annotation platform.

When to use it

• Preparing Kaggle time series datasets for machine learning or forecasting
• Standardizing Energy, Manufacturing, or Climate CSVs before annotation
• When you need a repeatable pipeline from data discovery to CLEAN status
• Uploading data with full metadata and unit consistency for downstream models
• Assigning consistent groups across targets and covariates for segmentation

Best practices

• Always run the provided pandas cleaning script before uploading the cleaned file
• Include precise metadata: dataset name, domain, source URL, and a concise description
• Verify at least one Time and one Target column are assigned before proceeding
• Set measurement units explicitly (kWh, kVarh, tCO2, ratio, seconds) to avoid ambiguity
• Apply group tags in bulk by selecting all target and covariate variables to ensure consistent segmentation

Example use cases

• Steel industry energy consumption dataset: clean CSV, set Usage_kWh as Target, assign kVarh and tCO2 units, and tag WeekStatus/Day_of_week/Load_Type across variables
• Climate CO2 time series: convert timestamps, fill missing emissions with median, set Target to CO2 and units to tCO2, assign regional group tags
• Manufacturing equipment telemetry: remove duplicates, fill sensor gaps, set Time column, mark covariates and group by production line and shift
• Rapidly ingest multiple Kaggle CSVs with the Kaggle CLI, clean via script, and bulk-configure metadata and groups for consistent ML-ready datasets