Adaptive models have poor performance
Adaptive models that are displayed in the adaptive models reports have low performance.
For more information about adaptive model reports, see Viewing summarized adaptive model reports.
Cause
The performance of a model is an indication of its predictive power. If the performance is low, the scores of the model have a low statistical correlation with the outcome that is to be predicted. The minimum value for the Coefficient of Concordance is 50.
There are a number of reasons why adaptive models might have low performance:
- Adaptive models miss some values for predictors.
- The values for predictors are not set.
- The values for parameterized predictors are not set.
- No active predictors are in the adaptive models.
- The adaptive models do not receive any responses or sufficient responses.
- The potential predictors available to the models are not correlated to the outcome.
Solution
- Depending on the cause, a system architect can perform the following
actions:
- Consider adding or removing predictors based on their performance.
- Change the settings in the adaptive model rule instance that is used to generate adaptive models.
- Try the solutions for the following issues:
Previous topic Adaptive Decision Manager does not capture responses for adaptive models Next topic No adaptive models are displayed in the adaptive model reports