Healthcare & Biomedicine Healthcare & Life Sciences use casesNACE Q86Binary ClassificationTarget: diabetes_diagnosisOpenML #37 ↗

Diabetes

Name: Pima Indians Diabetes Dataset (OpenML ID: 37)
Creator: National Institute of Diabetes and Digestive and Kidney Diseases
License: https://creativecommons.org/publicdomain/zero/1.0/

Clinical diagnostic benchmark comprising 768 patient records designed to predict the onset of diabetes mellitus in high-risk adult females based on diagnostic and metabolic biomarkers.

✨ Try in Carla Install Chrome Extension ↗100% In-Browser WebGPU • Zero Cloud Upload

768Records (Rows)

8Predictive Features

8 / 1Numeric / Categorical

0.0%Missing Value Ratio

0.16321 - AUC Error

1.16sWebGPU Mean Split

BUSINESS CONTEXT & OBJECTIVEClinical Diagnostics & Preventative Healthcare (Q86)

Business Objective: Diabetes

01Business Context

Primary care physicians, metabolic health clinics, and health insurance providers seek early-stage risk stratifications to identify individuals at elevated risk of Type 2 diabetes before irreversible vascular or organ damage occurs.

02Analytical Objective

Accurately predict binary diabetes onset (tested_positive vs. tested_negative) utilizing non-invasive demographic factors and routine clinical lab measurements.

03Economic & Decision Impact

Enables proactive preventative interventions (lifestyle programs, Metformin therapy) reducing long-term chronic care expenditure by up to 60%. Mitigates severe false negatives (delayed diagnosis leading to neuropathy, nephropathy, and cardiovascular events) while controlling false positives to avoid unnecessary clinical anxiety and second-line diagnostic costs.

ML Benchmark Narrative

On this classic 768-row tabular benchmark, optimized tree ensembles (LightGBM, XGBoost, Random Forest) traditionally achieve ROC-AUC scores between 0.82 and 0.84, often hampered by noisy zero-value physiological readings. TabICL and Carla in-context foundation models achieve competitive zero-shot discrimination (AUC ~0.83) instantly without hyperparameter tuning, handling sparse physiological representations directly in-context.

Source Origin:National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) / UCI Machine Learning Repository / OpenML

Creator:Vincent Sigillito (Johns Hopkins University Applied Physics Laboratory) & NIDDK researchers (1990)

License:Open Data / Public Domain (CC BY 4.0 compatible)

Reference:Paper / Repo ↗

FEATURE SPECIFICATION & DATA DICTIONARY

Target & Feature Column Definitions

Exhaustive business definitions, measurement units, roles, and target variables across all 9 columns.

Col	Variable / Feature Name	Role	Type	Unit / Scale	Description & Business Meaning
A	`diabetes_diagnosis`Diabetes Diagnosis Outcome	TARGET	CAT	—	Binary diagnostic indicator reflecting whether the patient was tested positive ('tested_positive') or negative ('tested_negative') for diabetes according to WHO criteria.
B	`pregnancy_count`Number of Pregnancies	Feature	NUM	count	Total number of times the patient has been pregnant; associated with gestational metabolic strain and insulin resistance.
C	`glucose_concentration_2h`2-Hour Plasma Glucose	Feature	NUM	mg/dL	Plasma glucose concentration measured 2 hours after an oral glucose tolerance test (OGTT). Key diagnostic indicator where values >= 200 mg/dL indicate diabetes.
D	`diastolic_blood_pressure`Diastolic Blood Pressure	Feature	NUM	mm Hg	Diastolic blood pressure reading; elevated values are correlated with metabolic syndrome and systemic vascular stiffness.
E	`triceps_skinfold_thickness`Triceps Skinfold Thickness	Feature	NUM	mm	Measurement of the fold of skin and subcutaneous fat over the triceps muscle, serving as an anthropometric proxy for peripheral body fat percentage.
F	`serum_insulin_2h`2-Hour Serum Insulin	Feature	NUM	mu U/ml	Serum insulin level measured 2 hours after glucose loading; assesses pancreatic beta-cell response and peripheral insulin sensitivity.
G	`body_mass_index`Body Mass Index (BMI)	Feature	NUM	kg/m²	Body mass index calculated as weight in kilograms divided by height in meters squared (kg/m²), indicating obesity levels.
H	`diabetes_pedigree_function`Diabetes Pedigree Function	Feature	NUM	score	Continuous genetic scoring function synthesizing family history of diabetes to estimate genetic risk predisposition.
I	`patient_age_years`Patient Age	Feature	NUM	years	Chronological age of the patient in years (all subjects are females >= 21 years old).

DATASET PREVIEW

Interactive Data Table

Explore rows, feature values, and target labels for Diabetes.

Download ParquetHF Viewer ↗

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

1.16s30 splits • 8 ensembles/split

Total Evaluation Runtime

34.86sIncludes warmup & sync

Evaluation Protocol

10×3 SplitsRepeated IID • 30 total

TabICLv2 Score

0.16321 - AUC Error

Evaluation Split	Train Rows	Test 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 / F1	512	256	1.17s	78.13%	0.8659	0.6543	0.7361	0.5889
R1 / F2	512	256	1.21s	76.95%	0.8340	0.6242	0.7206	0.5506
R1 / F3	512	256	1.19s	75.39%	0.8187	0.6441	0.6477	0.6404
R2 / F1	512	256	1.17s	75.00%	0.8275	0.5949	0.6912	0.5222
R2 / F2	512	256	1.24s	78.91%	0.8754	0.6538	0.7612	0.5730
R2 / F3	512	256	1.15s	73.44%	0.8151	0.5952	0.6329	0.5618
R3 / F1	512	256	1.15s	75.78%	0.8240	0.6517	0.6591	0.6444
R3 / F2	512	256	1.15s	75.78%	0.8483	0.6026	0.7015	0.5281
R3 / F3	512	256	1.21s	75.78%	0.8365	0.5921	0.7143	0.5056
R4 / F1	512	256	1.14s	76.56%	0.8401	0.6296	0.7083	0.5667
R4 / F2	512	256	1.16s	71.88%	0.8116	0.5500	0.6197	0.4944
R4 / F3	512	256	1.21s	78.13%	0.8498	0.6667	0.7089	0.6292
R5 / F1	512	256	1.15s	73.05%	0.8138	0.5965	0.6296	0.5667
R5 / F2	512	256	1.15s	78.13%	0.8517	0.6627	0.7143	0.6180
R5 / F3	512	256	1.14s	78.13%	0.8541	0.6456	0.7391	0.5730
R6 / F1	512	256	1.14s	78.13%	0.8662	0.6500	0.7429	0.5778
R6 / F2	512	256	1.19s	73.44%	0.8018	0.5952	0.6329	0.5618
R6 / F3	512	256	1.13s	75.39%	0.8321	0.5987	0.6912	0.5281
R7 / F1	512	256	1.14s	76.95%	0.8407	0.6509	0.6962	0.6111
R7 / F2	512	256	1.14s	76.56%	0.8276	0.6296	0.6986	0.5730
R7 / F3	512	256	1.15s	76.56%	0.8593	0.6104	0.7231	0.5281
R8 / F1	512	256	1.17s	75.00%	0.8205	0.6145	0.6711	0.5667
R8 / F2	512	256	1.15s	76.95%	0.8456	0.6335	0.7083	0.5730
R8 / F3	512	256	1.16s	77.34%	0.8458	0.6184	0.7460	0.5281
R9 / F1	512	256	1.15s	76.17%	0.8274	0.6514	0.6706	0.6333
R9 / F2	512	256	1.15s	75.78%	0.8495	0.5974	0.7077	0.5169
R9 / F3	512	256	1.15s	74.61%	0.8418	0.5963	0.6667	0.5393
R10 / F1	512	256	1.17s	76.95%	0.8420	0.6335	0.7183	0.5667
R10 / F2	512	256	1.15s	72.66%	0.8022	0.5732	0.6267	0.5281
R10 / F3	512	256	1.15s	76.17%	0.8423	0.6347	0.6795	0.5955
Mean ± Std	—	—	1.16s	75.99% ± 1.75%	0.837 ± 0.018	0.622 ± 0.028	0.692	0.566

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.

Rank	Algorithm / Model	Model Family	Runtime	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
#1	AttributeSelectedClassifier HoeffdingTree	Decision Tree	Python	77.34%	0.8190	0.7665	OpenML #577230 ↗
#2	ClassificationViaRegression M5P	Decision Tree	Python	77.34%	0.8304	0.7656	OpenML #578402 ↗
#3	AttributeSelectedClassifier SMO PolyKernel	Support Vector Machine (SVM)	Python	77.08%	0.7132	0.7592	OpenML #575789 ↗
#4	AttributeSelectedClassifier Bagging JRip	Bagging Ensemble	Python	77.08%	0.7984	0.7615	OpenML #576370 ↗
#5	AttributeSelectedClassifier Bagging LMT	Bagging Ensemble	Python	77.08%	0.8248	0.7666	OpenML #576534 ↗
#6	AttributeSelectedClassifier LMT	Machine Learning Model	Python	77.08%	0.8212	0.7622	OpenML #577469 ↗
#7	RandomForest	Random Forest	Python	76.95%	0.8312	0.7679	OpenML #573702 ↗
#8	Bagging REPTree	Decision Tree	Python	76.95%	0.8270	0.7648	OpenML #574537 ↗
#9	KernelLogisticRegression RBFKernel	Logistic / Linear Model	Python	76.82%	0.8381	0.7618	OpenML #323386 ↗
#10	SMO PolyKernel	Support Vector Machine (SVM)	Python	76.82%	0.7129	0.7577	OpenML #550234 ↗
#11	TabPFN v3	Tabular Foundation Model	PyTorch/Python	76.73%	0.8393	0.6354	PriorLabs/tabpfn ↗
#12	EXAONE Tabular	Tabular Foundation Model	PyTorch/Python	76.50%	0.8362	0.6332	LGAI-Research/EXAONE-Tabular ↗
#13	Google TabFM v1.0	Tabular Foundation Model	PyTorch/Python	76.42%	0.8377	0.6335	google-research/tabfm ↗
#14	nanotabicl Vanilla	Tabular Foundation Model	PyTorch/Python	76.20%	0.8368	0.6252	soda-inria/nanotabicl ↗
#15	Streaming nanotabicl	Tabular Foundation Model	PyTorch/Python	76.16%	0.8368	0.6248	soda-inria/nanotabicl ↗
#16	Full TabICLv2 (PyTorch Reference)	Tabular Foundation Model	PyTorch/Python	76.16%	0.8368	0.6248	soda-inria/tabicl ↗
#17	Carla Engine (TabICLv2 WebGPU)	Tabular Foundation Model	Browser	75.99%	0.8370	0.6217	100% In-Browser

STATISTICAL DISTRIBUTIONS

Variable Schema & Summary Distributions

Observed numerical ranges, category cardinalities, missing rates, and sample values across 768 rows.

Col	Variable Name	Type	Distinct	Summary Stats / Distribution	Sample Values
A	`diabetes_diagnosis`TARGET	CAT	2	2 cats: tested_negative, tested_positive	`tested_positivetested_negativetested_positive`
B	`pregnancy_count`	NUM	17	[`0`, `17`] μ=3.8 σ=3.4	`618`
C	`glucose_concentration_2h`	NUM	136	[`0`, `199`] μ=120.9 σ=32.0	`14885183`
D	`diastolic_blood_pressure`	NUM	47	[`0`, `122`] μ=69.1 σ=19.3	`726664`
E	`triceps_skinfold_thickness`	NUM	51	[`0`, `99`] μ=20.5 σ=15.9	`35290`
F	`serum_insulin_2h`	NUM	186	[`0`, `846`] μ=79.8 σ=115.2	`000`
G	`body_mass_index`	NUM	248	[`0`, `67.1`] μ=32.0 σ=7.9	`33.626.623.3`
H	`diabetes_pedigree_function`	NUM	517	[`0.078`, `2.42`] μ=0.5 σ=0.3	`0.6270.3510.672`
I	`patient_age_years`	NUM	52	[`21`, `81`] μ=33.2 σ=11.8	`503132`

FREQUENTLY ASKED QUESTIONS

Frequently Asked Questions: Diabetes

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

What is the Diabetes (Pima Indians) dataset used for?

The Diabetes dataset is a standard diagnostic machine learning benchmark containing 768 patient cases used to train and evaluate classification algorithms for predicting the onset of Type 2 diabetes based on 8 clinical and metabolic 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, allowing instant scoring without manual model training or backend infrastructure.

What machine learning models perform best on the Diabetes dataset?

Gradient boosting frameworks like LightGBM, XGBoost, and CatBoost alongside in-context foundation models like TabICL typically achieve the strongest classification performance, yielding ROC-AUC scores in the 0.82-0.84 range.

LIVE EVALUATION IN GOOGLE SHEETS

Test TabICLv2 on Diabetes 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 Diabetes in Carla Install Chrome Extension ↗

Provenance & Attribution

https://www.jair.org/index.php/jair/article/view/10129

License: PublicData Source: OpenMLOpenML Page: https://www.openml.org/d/37

Raw Data Downloads

Download Parquet OpenML Task ↗Google Sheet ↗

← Previous DatasetCredit GFinance & Banking All Datasets Next Dataset →Employee SalariesHuman Resources & Workforce Analytics