What it is
569 fine-needle aspirate (FNA) samples of breast masses. Each row is one sample, described by 30 numeric features computed from a digitised image of the cell nuclei: the mean, standard error and worst (largest) value of ten measurements (radius, texture, perimeter, area, smoothness, compactness, concavity, concave points, symmetry, fractal dimension). The diagnosis column is M (malignant, 212) or B (benign, 357).
How to use it
A common benchmark for classification. Hold out a test set before tuning; report sensitivity and specificity, not only accuracy. Validation runs against the original values can be requested from the institute.
Source and changes
University of Wisconsin creators, hosted by the UCI Machine Learning Repository (doi:10.24432/C5DW2B). We added descriptive column names and removed the original record ID; all samples and values are otherwise unchanged.
Limitations
Single centre, early 1990s imaging. Not for clinical decisions.