Explore what signals heart disease
- The problem
- Many routine measurements relate to heart disease, and it is hard to see which ones carry the most signal and how they combine.
- What the demo shows
- Compare patients with and without diagnosed heart disease on each measure, choose inputs for a simple model, and see how well it separates the two groups.
- Why it matters
- Shows, on a well-known public dataset, how a transparent model is built and checked before anyone relies on it.
Patients
0
Cleveland subsetWith heart disease
0
Angiographic narrowing over 50%Median age
0
YearsNo heart diseaseHeart disease
Choose model inputs
A logistic regression is fitted in your browser on 70% of patients and checked on the other 30%. Missing values (6 patients) are filled with the most common value.
Area under the ROC curve, held-out 30%
0.00
How each input moves the odds
Odds ratio for a one standard deviation increase (or for yes versus no). Above 1 raises the modeled odds of heart disease; below 1 lowers them. Associations in this sample, not causes.
What-if profile
Adjust a sample profile. The estimate uses the model and inputs chosen on the previous tab.
Modeled probability of heart disease
0%
Contribution of each input
Interactive demo · public research data (UCI Heart Disease, Cleveland), not customer results.