Penn Machine Learning Benchmarks (PMLB) is a large collection of curated benchmark datasets for evaluating and comparing supervised machine learning algorithms. These datasets cover a broad range of applications including binary/multi-class classification and regression problems as well as combinations of categorical, ordinal, and continuous features.

Summary statistics

In the interactive plotly chart below, each dot represents a dataset colored based on its associated task (classification vs. regression). In log scale, the x and y axis shows the number of observations and features respectively. Please click on the legend to hide/show the groups of datasets. Click on each dot to access the dataset’s pandas-profiling report.

Note: If a dataset has more than 20 features, we randomly chose 20 to be displayed in its profiling report. Therefore, please disregard the Number of variables in the corresponding report and, instead, use the correct n_features in the chart and table below.

Browse, sort, filter and search the complete table of summary statistics below.

  • Click on the dataset’s name to access its pandas-profiling report.

  • Click on the GitHub Octocat to access its metadata.

  • To filter, please type in the box at the bottom of each numeric column in the format low ... high. For example, if you want to see all classification datasets with 80 to 100 observations, select classification at the bottom of Task and type 80 ... 100 at the bottom of the n_observations column.

The complete table of dataset characteristics is also available for download. Please note, in our documentation, a feature is considered:

  • “binary” if it is of type integer and has 2 unique values (equivalent to pandas profiling’s “boolean”)
  • “categorical” if it is of type integer and has more than 2 unique values (equivalent to pandas profiling’s “categorical”)
  • “continuous” if it is of type float (equivalent to pandas profiling’s “numeric”).

Dataset format

All datasets are stored in a common format:

  • First row is the column names
  • Each following row corresponds to one observation of the data
  • The dependent variable/endpoint/outcome column is named target
  • All columns are tab (\t) separated
  • All files are compressed with gzip to conserve space

Citing PMLB

If you use PMLB in a scientific publication, please consider citing one of the following papers:

Joseph D. Romano, Le, Trang T., William La Cava, John T. Gregg, Daniel J. Goldberg, Praneel Chakraborty, Natasha L. Ray, Daniel Himmelstein, Weixuan Fu, and Jason H. Moore. PMLB v1.0: an open source dataset collection for benchmarking machine learning methods. arXiv preprint arXiv:2012.00058 (2020).

  title={PMLB v1.0: an open source dataset collection for benchmarking machine learning methods},
  author={Romano, Joseph D and Le, Trang T and La Cava, William and Gregg, John T and Goldberg, Daniel J and Chakraborty, Praneel and Ray, Natasha L and Himmelstein, Daniel and Fu, Weixuan and Moore, Jason H},
  journal={arXiv preprint arXiv:2012.00058v2},

Olson, Randal S., William La Cava, Patryk Orzechowski, Ryan J. Urbanowicz, and Jason H. Moore. PMLB: a large benchmark suite for machine learning evaluation and comparison. BioData mining 10, no. 1 (2017): 1-13. BioData Mining 10, page 36.

BibTeX entry:

    author="Olson, Randal S. and La Cava, William and Orzechowski, Patryk and Urbanowicz, Ryan J. and Moore, Jason H.",
    title="PMLB: a large benchmark suite for machine learning evaluation and comparison",
    journal="BioData Mining",

Support for PMLB

PMLB was developed in the Computational Genetics Lab at the University of Pennsylvania with funding from the NIH under grant AI117694, LM010098 and LM012601. We are grateful for the support of the NIH and the University of Pennsylvania during the development of this project.