Meta-level modeling

Meta-level modeling is building a model based on predictions or results from another model. In the previous example, we saw how to create propensity scores for our Chaid model. In this section, we will see how you can extract results from the Chaid model and feed them into a Neural Net model, and this will enable us to improve the results from a Neural Net model.

To do this, follow these steps:

1. Connect the partition node to the Neural Net node from the modeling palette.
2. Run the Neural Net model by changing the Random seed to 5000 from the Advanced options under the Build tab and click on Run. Now connect the generated Chaid model to the generated Neural Net model.
3. Now, use the Analysis node to see the level of accuracy of these models.
4. You will get the following results:

We can see that the accuracy of both the models is somewhat similar, so now we will...