CARLA HQ
Add to Chrome
Healthcare & Life Sciences Healthcare & Life Sciences use casesNACE Q86Binary ClassificationTarget: diagnosis_classOpenML #15 ↗

Breast W

Clinical cytology benchmark containing 699 patient fine needle aspirate (FNA) biopsy records for binary classification of breast tumors into benign or malignant diagnoses based on nuclear morphological characteristics.

✨ Try in CarlaInstall Chrome Extension ↗100% In-Browser WebGPU • Zero Cloud Upload
699Records (Rows)
9Predictive Features
9 / 1Numeric / Categorical
0.2%Missing Value Ratio
0.00521 - AUC Error
1.13sWebGPU Mean Split
BUSINESS CONTEXT & OBJECTIVEMedical Diagnostics & Clinical Oncology (Q86)

Business Objective: Breast W

01Business Context

Early, accurate diagnosis of breast neoplasms from fine needle aspiration (FNA) biopsies is critical in clinical pathology workflows to differentiate non-cancerous lesions from invasive carcinomas without requiring invasive surgical biopsies.

02Analytical Objective

Predict whether a breast mass biopsy sample is benign or malignant using digitized microscopic nuclear morphometric features graded on an ordinal 1-10 clinical scale.

03Economic & Decision Impact

Reduces turnaround time for biopsy triaging and eliminates unnecessary surgical interventions for benign cases. False negatives carry life-threatening diagnostic delays and elevated mortality risks, while false positives result in severe patient psychological distress and costly unnecessary surgical procedures.

ML Benchmark Narrative

Standard supervised tree ensembles like XGBoost and Random Forest reliably achieve 96-97% classification accuracy and strong ROC-AUC on this classic clinical pathology dataset. Zero-shot in-context tabular foundation models (TabICL / Carla) match these competitive diagnostic benchmarks instantly without requiring hyperparameter tuning, cross-validation pipelines, or separate model training phases.

Source Origin:University of Wisconsin Hospitals, Madison / UCI Machine Learning Repository / OpenML
Creator:Dr. William H. Wolberg, W. Nick Street, and Olvi L. Mangasarian (1992)
License:CC BY 4.0 / Public Domain (UCI Open Data)
Reference:Paper / Repo ↗
FEATURE SPECIFICATION & DATA DICTIONARY

Target & Feature Column Definitions

Exhaustive business definitions, measurement units, roles, and target variables across all 10 columns.

ColVariable / Feature NameRoleTypeUnit / ScaleDescription & Business Meaning
Adiagnosis_classDiagnostic Pathology ClassTARGETCAT—
Final histological and cytological diagnostic determination: benign non-cancerous tumor or malignant carcinoma.
Bclump_thicknessClump ThicknessFeatureNUMscale (1-10)
Grading of whether cells are grouped in mono- or multi-layers; higher values indicate multilayered cellular grouping typical of malignancy (graded 1 to 10).
Ccell_size_uniformityCell Size UniformityFeatureNUMscale (1-10)
Assessment of consistency in cell size across the biopsy sample; high variation indicates anaplasia and malignant progression (graded 1 to 10).
Dcell_shape_uniformityCell Shape UniformityFeatureNUMscale (1-10)
Evaluation of nuclear shape consistency and regular cellular geometry versus pleomorphic irregular shapes (graded 1 to 10).
Emarginal_adhesionMarginal AdhesionFeatureNUMscale (1-10)
Degree of cohesion between neighboring cells; loss of cell adhesion is a hallmark of malignant invasive potential (graded 1 to 10).
Fsingle_epithelial_cell_sizeSingle Epithelial Cell SizeFeatureNUMscale (1-10)
Degree of cellular enlargement in single epithelial cells; significantly enlarged cells correlate with malignancy (graded 1 to 10).
Gbare_nucleiBare Nuclei CountFeatureNUMscale (1-10)
Presence of cell nuclei unencapsulated by cytoplasm; elevated bare nuclei counts are strongly associated with benign vs malignant status (graded 1 to 10).
Hbland_chromatinBland ChromatinFeatureNUMscale (1-10)
Nuclear chromatin texture uniformity; fine/bland texture indicates benignity while coarse, clotted chromatin indicates malignancy (graded 1 to 10).
Inormal_nucleoliNormal Nucleoli ProminenceFeatureNUMscale (1-10)
Prominence and hypertrophy of nucleoli within cell nuclei; large, prominent nucleoli suggest active malignancy (graded 1 to 10).
Jmitoses_countMitotic ActivityFeatureNUMscale (1-10)
Rate and frequency of cell division (mitotic figures) observed under high-power microscopy (graded 1 to 10).
DATASET PREVIEW

Interactive Data Table

Explore rows, feature values, and target labels for Breast W.

Download ParquetHF Viewer ↗
Loading dataset...
LOCAL EXECUTION TELEMETRY

TabICLv2 WebGPU Benchmark Results

Complete performance metrics on persisted IID or non-IID evaluation splits, executed 100% locally in the browser sandbox.

Mean Split Latency?Mean wall-clock execution time per persisted evaluation split running 100% locally via WebGPU inside Chrome.Explore Latency Guide →
1.13s60 splits • 8 ensembles/split
Total Evaluation Runtime?Total wall-clock duration across all selected splits, including in-context encoding, WebGPU shader execution, and result aggregation.Explore Split Protocol →
67.71sIncludes warmup & sync
Evaluation Protocol?Precomputed row indices ensure the browser and Python runners evaluate identical out-of-sample observations without leakage.Explore Split Protocol →
20×3 SplitsRepeated IID • 60 total
TabICLv2 Score?Primary task performance score (1 - AUC Error) achieved by TabICLv2 across out-of-sample evaluation splits.Explore Metric Formula →
0.00521 - AUC Error
Evaluation SplitTrain RowsTest Rows
Split Duration?Wall-clock inference time taken for this evaluation split running 100% locally via WebGPU.Learn more →
Accuracy?Proportion of correct test predictions across all classes.Google ML Guide & Details →
ROC-AUC?Area under the ROC curve evaluating ranking capability across all thresholds.Google ML Guide & Details →
F1 Score?Harmonic mean of Precision and Recall, robust against class imbalance.Learn more →
Precision?Proportion of predicted positives that were actually positive.Google ML Guide & Details →
Recall?Proportion of actual positive cases successfully captured.Google ML Guide & Details →
R1 / F14662331.14s98.28%0.99840.97560.95241.0000
R1 / F24662331.13s97.00%0.99440.95650.95060.9625
R1 / F34662331.11s96.57%0.99330.95060.95060.9506
R2 / F14662331.12s96.14%0.99060.94410.93830.9500
R2 / F24662331.11s96.57%0.99520.95120.92860.9750
R2 / F34662331.15s97.00%0.99550.95650.96250.9506
R3 / F14662331.14s95.71%0.98910.94120.88891.0000
R3 / F24662331.09s98.28%0.99590.97500.97500.9750
R3 / F34662331.13s97.00%0.99570.95600.97440.9383
R4 / F14662331.14s96.14%0.99340.94480.92770.9625
R4 / F24662331.10s97.00%0.99540.95600.96200.9500
R4 / F34662331.12s97.85%0.99680.96970.95240.9877
R5 / F14662331.11s98.28%0.99970.97441.00000.9500
R5 / F24662331.10s97.85%0.99540.96890.96300.9750
R5 / F34662331.15s95.28%0.98610.93490.89770.9753
R6 / F14662331.12s97.42%0.99720.96250.96250.9625
R6 / F24662331.13s95.71%0.98980.93900.91670.9625
R6 / F34662331.13s97.42%0.99650.96300.96300.9630
R7 / F14662331.12s98.28%0.99970.97560.95241.0000
R7 / F24662331.12s96.57%0.99220.95060.93900.9625
R7 / F34662331.11s96.57%0.99350.95120.93980.9630
R8 / F14662331.12s97.42%0.99460.96250.96250.9625
R8 / F24662331.12s97.00%0.99570.95650.95060.9625
R8 / F34662331.09s95.71%0.99450.93900.92770.9506
R9 / F14662331.11s96.14%0.99570.94410.93830.9500
R9 / F24662331.12s96.57%0.99130.95060.93900.9625
R9 / F34662331.10s98.28%0.99850.97590.95291.0000
R10 / F14662331.12s97.00%0.99150.95650.95060.9625
R10 / F24662331.13s97.85%0.99830.96970.94121.0000
R10 / F34662331.13s97.42%0.99550.96340.95180.9753
R11 / F14662331.11s95.28%0.98910.93170.92590.9375
R11 / F24662331.10s96.14%0.99610.94550.91760.9750
R11 / F34662331.12s98.71%0.99820.98140.98750.9753
R12 / F14662331.15s97.00%0.99580.95650.95060.9625
R12 / F24662331.28s96.14%0.99560.94340.94940.9375
R12 / F34662331.23s97.00%0.99500.95860.92051.0000
R13 / F14662331.16s97.85%0.99670.96820.98700.9500
R13 / F24662331.10s96.57%0.99240.95060.93900.9625
R13 / F34662331.24s96.57%0.99340.95180.92940.9753
R14 / F14662331.14s96.57%0.99580.95180.91860.9875
R14 / F24662331.14s97.85%0.99800.96930.95180.9875
R14 / F34662331.17s95.28%0.98970.93080.94870.9136
R15 / F14662331.12s96.14%0.99200.94550.91760.9750
R15 / F24662331.11s98.71%0.99940.98110.98730.9750
R15 / F34662331.11s97.85%0.99290.96970.95240.9877
R16 / F14662331.11s96.57%0.99580.95180.91860.9875
R16 / F24662331.14s96.57%0.99490.94870.97370.9250
R16 / F34662331.13s97.00%0.99340.95810.93020.9877
R17 / F14662331.11s96.14%0.99490.94340.94940.9375
R17 / F24662331.14s97.00%0.99670.95760.92940.9875
R17 / F34662331.12s96.14%0.99510.94410.95000.9383
R18 / F14662331.11s96.57%0.99630.94940.96150.9375
R18 / F24662331.10s96.57%0.99130.95180.91860.9875
R18 / F34662331.14s97.00%0.99630.95650.96250.9506
R19 / F14662331.18s98.71%0.99900.98160.96391.0000
R19 / F24662331.11s95.71%0.99530.93830.92680.9500
R19 / F34662331.12s95.71%0.98900.93830.93830.9383
R20 / F14662331.13s97.42%0.99780.96250.96250.9625
R20 / F24662331.08s96.14%0.99460.94480.92770.9625
R20 / F34662331.12s97.00%0.99120.95710.95120.9630
Mean ± Std——1.13s96.90% ± 0.89%0.995 ± 0.0030.956 ± 0.0130.9460.966
UNIFIED BENCHMARK LEADERBOARD

Foundation Models vs Traditional ML Baselines

Side-by-side evaluation of our in-browser Carla engine (WebGPU), open-source tabular foundation models, and OpenML baselines.

RankAlgorithm / ModelModel FamilyRuntime
Predictive Accuracy?Proportion of correct test predictions across cross-validation splits.Google ML Guide & Details →
ROC-AUC?Area under the ROC curve evaluating ranking capability.Learn more →
F-Measure?Harmonic mean of precision and recall.Learn more →
Reference / Repo
#1AttributeSelectedClassifier PrincipalComponents Ranker J48Decision TreePython97.14%0.97180.9716OpenML #251026 ↗
#2A1DEMachine Learning ModelPython97.14%0.98820.9715OpenML #298573 ↗
#3AttributeSelectedClassifier A1DEMachine Learning ModelPython97.14%0.98820.9715OpenML #575891 ↗
#4EXAONE TabularTabular Foundation ModelPyTorch/Python97.00%0.99460.9572LGAI-Research/EXAONE-Tabular ↗
#5BayesNet K2Naive BayesPython97.00%0.98900.9701OpenML #429 ↗
#6RandomForestRandom ForestPython97.00%0.99120.9700OpenML #573524 ↗
#7AttributeSelectedClassifier RandomForestRandom ForestPython97.00%0.99170.9700OpenML #575708 ↗
#8AttributeSelectedClassifier BayesNetNaive BayesPython97.00%0.98900.9701OpenML #575734 ↗
#9BayesNetNaive BayesPython97.00%0.98900.9701OpenML #577577 ↗
#10Google TabFM v1.0Tabular Foundation ModelPyTorch/Python96.97%0.99450.9567google-research/tabfm ↗
#11nanotabicl VanillaTabular Foundation ModelPyTorch/Python96.92%0.99480.9560soda-inria/nanotabicl ↗
#12Streaming nanotabiclTabular Foundation ModelPyTorch/Python96.91%0.99480.9558soda-inria/nanotabicl ↗
#13Full TabICLv2 (PyTorch Reference)Tabular Foundation ModelPyTorch/Python96.91%0.99480.9558soda-inria/tabicl ↗
#14Carla Engine (TabICLv2 WebGPU)Tabular Foundation ModelBrowser96.90%0.99470.9556100% In-Browser
#15RandomRulesMachine Learning ModelPython96.85%0.99040.9686OpenML #54746 ↗
#16TabPFN v3Tabular Foundation ModelPyTorch/Python96.84%0.99440.9547PriorLabs/tabpfn ↗
#17SPegasosMachine Learning ModelPython96.71%0.96600.9672OpenML #573967 ↗
STATISTICAL DISTRIBUTIONS

Variable Schema & Summary Distributions

Observed numerical ranges, category cardinalities, missing rates, and sample values across 699 rows.

ColVariable NameTypeMissingDistinctSummary Stats / DistributionSample Values
Adiagnosis_classTARGETCAT022 cats: benign, malignantbenignbenignbenign
Bclump_thicknessNUM010[1, 10] μ=4.4 σ=2.8553
Ccell_size_uniformityNUM010[1, 10] μ=3.1 σ=3.0141
Dcell_shape_uniformityNUM010[1, 10] μ=3.2 σ=3.0141
Emarginal_adhesionNUM010[1, 10] μ=2.8 σ=2.9151
Fsingle_epithelial_cell_sizeNUM010[1, 10] μ=3.2 σ=2.2272
Gbare_nucleiNUM16 (2.29%)10[1, 10] μ=3.5 σ=3.61102
Hbland_chromatinNUM010[1, 10] μ=3.4 σ=2.4333
Inormal_nucleoliNUM010[1, 10] μ=2.9 σ=3.1121
Jmitoses_countNUM09[1, 10] μ=1.6 σ=1.7111
FREQUENTLY ASKED QUESTIONS

Frequently Asked Questions: Breast W

Common questions regarding the Breast W dataset, machine learning task formulations, and in-browser tabular inference.

What is the Breast Cancer Wisconsin (Original) dataset used for?

The Breast Cancer Wisconsin (Original) dataset is a standard biomedical benchmark used to train and evaluate classification algorithms in distinguishing between benign and malignant breast neoplasms based on 9 microscopic cytological features.

Can I run zero-server predictions on this dataset in Google Sheets?

Yes, Carla HQ enables zero-server in-context tabular prediction directly in spreadsheets using TabICL foundation models without needing code, APIs, or dedicated GPU pipelines.

What machine learning models perform best on the Breast Cancer Wisconsin dataset?

Gradient boosted decision trees (XGBoost, LightGBM), Support Vector Machines (SVM), and tabular foundation models like TabICL routinely achieve top-tier diagnostic accuracy (96-97%+) and high area under the ROC curve (ROC-AUC) on this benchmark.

LIVE EVALUATION IN GOOGLE SHEETS

Test TabICLv2 on Breast W Yourself

Open the pre-loaded Google Sheet and let Carla configure the target and task for local, zero-cloud tabular machine learning.

Step 1

Launch Carla

Click once to open the spreadsheet and Carla side panel together.

Step 2

Review the Setup

Carla selects the dataset target and task from this page automatically.

Step 3

Evaluate & Predict

Run predictions and compute metrics with zero server uploads.

✨ Try Breast W in CarlaInstall Chrome Extension ↗

Provenance & Attribution

Dr. William H. Wolberg, University of Wisconsin Hospitals, Madison.

License: PublicData Source: OpenMLOpenML Page: https://www.openml.org/d/15

Raw Data Downloads

Download ParquetOpenML Task ↗Google Sheet ↗
All Datasets