Analytics
Logistic Regression
Fits a logistic model that gives the probability of a yes/no outcome from other columns, plus a 1/0 prediction at the threshold you choose.
When to use it
Use it for churn, conversion, default or any other yes/no question, when you want a probability rather than a hard rule.
In a pipeline
File InputTrain/Test SplitLogistic RegressionExport CSV
Settings
| Setting | Type | Required | Default or hint |
|---|---|---|---|
| Predict (yes/no column) | select | Yes | Select column… |
| Value that means yes | text | — | Default: 1, true, yes or y |
| From columns (features) | select | Yes | Pick the columns that explain it |
| Yes from probability | number | — | 0.5 |
| Probability column name | text | — | Default: Probability |
| Prediction column name | text | — | Default: Predicted |
| Add a Model column (JSON for Score Model) | checkbox | — | false |
Example
Churn probability from support tickets
- Predict (yes/no column)=Churned
- From columns (features)=Tickets
| Customer | Tickets | Churned |
|---|---|---|
| A | 0 | no |
| B | 1 | no |
| C | 2 | yes |
| D | 3 | no |
| E | 4 | yes |
| F | 5 | yes |
| Customer | Tickets | Churned | Probability | Predicted |
|---|---|---|---|---|
| A | 0 | no | 0.0459 | 0 |
| B | 1 | no | 0.1393 | 0 |
| C | 2 | yes | 0.3528 | 0 |
| D | 3 | no | 0.6473 | 1 |
| E | 4 | yes | 0.8607 | 1 |
| F | 5 | yes | 0.9541 | 1 |
Worth knowing
- Yes is 1, true, yes or y and no is 0, false, no or n — or type the Value that means yes, and every other value counts as no.
- Fitted by Newton's method with a tiny ridge, so even perfectly separated data gives an answer; the note reports training accuracy.
- Lower Yes from probability to catch more of the yes cases at the cost of more false alarms.
- The model is JSON in the notes (or the Model column) for Score Model; nothing is stored.