Will someone who sees the ad buy?
Enter an age and an estimated salary. The model predicts how likely this person is to buy the advertised product, from what it learned on a teaching dataset of 400 people.
In years, for example 42
For example 85,000
Purchase probability
–
Who should see the ad?
People with a probability above the threshold see the ad. Lower it and you reach more buyers, but more showings are wasted.
| –Right targetShown, bought | –Wasted showingShown, did not buy |
| –Missed buyerNot shown, would have bought | –Right skipNot shown, would not have bought |
- Precision
- –
- How many who saw the ad bought
- Recall
- –
- How many buyers were reached
- F1
- –
These numbers come from predictions made by a model that never saw that person in training (10-fold cross-validation).
How it works
Plain logistic regression can only draw a straight boundary and falls behind. Adding the squares and the product of age and salary bends the boundary, and the simple model matches the complex ones. Switch between models on the map to see the difference.
Data and limits
The data is a teaching dataset of 400 people widely used in machine learning courses: gender, age, estimated salary and whether they bought the advertised product. It is not data from a real campaign; the results show how the methods work.
Gender is left out of the model: it adds nothing to the prediction here, and using it to target ads raises a fairness question.
The age and salary you enter are not sent or stored anywhere; the prediction runs in your browser.