Electricity: Subsampled Energy Consumption Data for ML Benchmarking

A 2000-row, 100-column subsample of a larger electricity dataset, created with a specific random seed for reproducibility. The subset includes up to 10 target classes and was generated using a stratified sampling method. This version was created by Eddie Bergman and is shared under a U.S. public domain license on OpenML.

Use Cases

• Benchmarking classification model performance on a controlled, reproducible subset of energy data.
• Testing feature selection and dimensionality reduction methods on a dataset with 100 columns.
• Comparing stratified versus non-stratified sampling strategies for model training.
• Developing educational examples for machine learning courses using a manageable, real-world dataset.

Strengths

• Controlled subsample size of 2000 rows and 100 columns, enabling faster experimentation.
• Explicitly documented generation parameters (seed=3, stratified sampling) for full reproducibility.
• Derived from a larger source dataset (OpenML ID 44120), suggesting a foundation in real-world measurements.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the original source dataset is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

OpenML, derived from dataset with ID 44120.

Collection Method

Algorithmic subsampling from a larger source dataset using a Python script with NumPy and scikit-learn.