PMLB

PMLB provides a repository of curated supervised machine learning benchmark datasets for standardized evaluation and comparison of algorithms.

Key Features:

• Curated benchmark datasets: Aggregates curated, well-studied public datasets intended for supervised machine learning evaluation.
• Supervised learning focus: Concentrates on datasets formatted for supervised tasks used in classification and regression benchmarking.
• Standardization for comparison: Presents datasets in a standardized form to support consistent evaluation and comparison of algorithms and statistical models.
• Large, diverse aggregation: Includes a broad collection of diverse public benchmark datasets consolidated in a single repository.
• Programmatic language interfaces: Provides programmatic access via Python and R interfaces for integration with analysis workflows.

Scientific Applications:

• Algorithm benchmarking: Enables systematic evaluation and comparison of supervised machine learning algorithms across multiple datasets.
• Method development and validation: Supports development, validation, and performance assessment of novel machine learning and statistical modeling methods.
• Comparative studies: Facilitates comparative studies that require standardized datasets to quantify algorithm performance differences.

Methodology:

Aggregates and curates public supervised learning datasets and exposes them via programmatic Python and R interfaces.