Electricity: Subsampled Machine Learning Benchmark with 10 Classes

Electricity_seed_2_nrows_2000_nclasses_10_ncols_100_stratify_True is a subsampled version of the 'electricity' dataset from OpenML, created by Eddie Bergman. It was generated with a random seed of 2, targeting 2000 rows, 100 columns, and 10 classes. The subsampling process included stratified sampling to preserve class distribution.

Use Cases

• Benchmarking classification model performance on a controlled, subsampled dataset.
• Testing feature selection methods on a dataset with 100 columns.
• Evaluating model behavior on a multi-class problem with 10 target classes.
• Comparing stratified versus non-stratified sampling techniques in data preprocessing.

Strengths

• Dataset creation parameters are explicitly documented: seed=2, nrows=2000, ncols=100, nclasses=10.
• Subsampling was performed with stratification, which likely preserves the original dataset's class balance.
• The source dataset (OpenML ID 44120) is a known benchmark for electricity-related forecasting.

Limitations

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

Provenance

Source

OpenML dataset 'electricity' (ID 44120), subsampled by Eddie Bergman.

Collection Method

Algorithmic subsampling using a Python script with NumPy and scikit-learn's train_test_split.