Modified on Mon, 27 Mar 2023 at 10:08 AM


Using a Post-Processor with a model combines applying a Calculator and a Chain Model on the same model into a single step. It picks up a model, applies a calculation operation on its inputs and/or outputs to create a new output and forms a wrapper around the old model.


Situations when a Post-Processor could be applied:

  • Convert an output to a different unit.
  • Convert a probability into a category. That way you could turn a regression model into a classification model.
  • Calculate a new output for which the mathematical relationship is known (e.g. calculate force if pressure and surface area is known).

The Post-Processor is limited to the features known to the model on which it is applied. Only those model’s inputs and outputs are available. No external information can be included in these operations. If that is required, the Calculator model with an additional Chain model step might be the better solution.

How to use

  • Select a Model. Only a single model can be selected.
  • Specify the Name of new output.
  • Select the Operation which should be performed to generate the output. The same options are available as for Quick Columns. Also, the same limitations apply to the custom code environment. Refer to the Quick Columns article for more details.
  • After having selected the operation you need to specify on which columns the Post-Processor model should operate. You can choose only from among the model inputs and outputs.
  • Give the resulting model a Name.
  • You can decide if the original model’s output(s) will be exposed or not in the Post-Processor model. If they should be exposed enable the option Include intermediate outputs in model's outputs. This behavior is similar to the Chain Model’s behavior.
  • Click Apply to create the new model.


Consider the following use case: based on an initial temperature and some other inputs a resulting temperature is predicted. The first attempt is to predict the absolute temperature value directly but the model is not achieving the required accuracy.

The model is changed to predict the temperature difference instead which works much better (ΔT = Tabs - Tinit). But what is actually needed is the absolute temperature value. To achieve this a Post-Processor could be applied to the model predicting the temperature difference

As the post-processor can work with all inputs and outputs of the underlying model it would be possible to use the initial temperature (model input) and add the predicted temperature difference (output) to calculate the absolute temperature value as new output. The final model (Mfinal) would be a wrapper around the initial model. It would have the same inputs as the original model and the new absolute temperature value as output. With the option Include intermediate outputs in model's outputs the original model’s output ΔT could also be exposed as model output.

Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select atleast one of the reasons
CAPTCHA verification is required.

Feedback sent

We appreciate your effort and will try to fix the article