Deploying the two AIs together (ML and rules)
ML is great, but this is a book mainly about AI business rules. Our first step to combine the two is to open and edit our PMML model in Azure, as in Figure 11.17. This PMML model was generated in the last step of our Python notebook. You may need to click on the Refresh button in Azure ML to see this newly generated model:
Figure 11.17 – PMML generated by the ML script
As you can see, PMML is an XML format that is largely human-readable. It also has the advantage of being an industry standard, so even though sklearn
and a Python notebook generated the model, we are able to import and execute the model in KIE. In general, the interoperability works smoothly—but a bit like assembling a cabinet from Ikea, sometimes a few gentle knocks from a mallet are needed to get everything into place.
In our case, there are two edits needed to the PMML file so that Kogito can work with it:
- On the second...