Machine Learning
Predicting Diabetes Using Machine Learning
Enter a patient's health measurements and a trained model estimates whether they are likely to have diabetes. Built end to end in Python, from raw data to a model that runs right here in your browser.
- 768Patient records used
- 8Measurements per patient
- –Accuracy on unseen patients
- –ROC AUC score
Explore the project
Try the predictor first, then look at the data and the thinking behind it.
Predict
Enter a patient's measurements and see the model's estimate.
Open predictorThe data
What each measurement means and how patients with diabetes differ.
See the dataThe method
Cleaning, training and testing, step by step, in plain language.
Read the methodThe results
How accurate it really is, including where it gets things wrong.
View resultsFrom records to a prediction
The whole pipeline in four moves.
-
Patient data
768 records, 8 measurements each
-
Clean and scale
Fix missing values, balance units
-
Train
Logistic regression learns the pattern
-
Predict
Runs instantly in your browser
See what the model says about a patient
It takes under a minute. Only glucose, BMI and age are required.