Lift Chart
Creates a lift chart. Additionally, a chart for the cumulative
percent of responses captured is shown. A lift chart is used to evaluate
a predictive model. The higher the lift (the difference between the "lift" line
and the base line), the better performs the predictive model.
The lift is the ratio between the results obtained with and without the
predictive model.
It is calculated as number of positive hits (e .g. responses) divided by
the average number of positives without model.
The data table must have a column containing probabilities and a nominal
column, containing the actual labels.
At first, the data is sorted by probability, divided into deciles,
then the actual labels are counted and the average rate is calculated.
Dialog Options
- Column containing true labels
- Nominal column containing the actual labels, e. g. if a person responded
- Response Label
- The label for a positive value (hit).
- Column containing score (probabilities)
- Numeric column containing the predicted score in probabilities of the model
- Interval width in %
- The width in which the data is separated before counting.
Ports
Output Ports
0 |
Data table sorted by probability |
Views
- Lift Chart
- The lift chart
This node is contained in KNIME Base Nodes
provided by KNIME GmbH, Konstanz, Germany.