Diabetes Prediction Models

Train SVM, KNN, Logistic Regression and Decision Tree on real patient data. Evaluate with F1, Precision, Recall, AUC and predict individual patient risk.

Get Started

Train All Models

One click trains SVM, KNN, Logistic Regression & Decision Tree on 614 samples

768
Patients
34.9%
Diabetic Rate
—
Best Accuracy
—
Best ROC-AUC
Train / Test Split
80 / 20
Drag to change how much data trains vs. tests. Too little test data → unreliable scores.
Hyperparameters
SVM C 1.0
KNN k 9
Tree Depth 5
Feature Selection (min 2)

Models not trained yet

Class Balance
Diabetic 35%Healthy 65%
Glucose Distribution
BMI Distribution

Model Performance

Train models to see results

SVM

SVM · RBF

—
Awaiting patient…
KNN

KNN · k=9

—
Awaiting patient…
LR

Logistic Reg.

—
Awaiting patient…
DTree

Decision Tree

—
Awaiting patient…
Model Comparison
Feature Radar

ROC Curves

Receiver Operating Characteristic — higher AUC = better discrimination between classes

The diagonal dashed line is a random classifier (AUC = 0.5). A perfect model hugs the top-left corner.

Confusion Matrix

Classification report on test set (154 samples)

Matrix
—True Pos
—False Pos
—False Neg
—True Neg
Classification Report
ClassPrecisionRecallF1Support

Learning Curves

Training vs cross-validation score as training set grows

Training scoreCV validation

Feature Importance

Risk factor rankings from Logistic Regression coefficients

#FeatureImportanceBarDiabetic AvgHealthy AvgΔ

Feature Scatter

Explore relationships between any two features

vs

Patient Risk Predictor

Enter patient data — all 4 trained models vote on the outcome

Patient Metrics
Prediction Result

Enter patient data and click Analyse Patient Risk

Prediction History

All past patient analyses

No analyses yet.