Summary

The topics we covered in this chapter were the idea that ML is as simple as drawing a line on a graph, with the learning piece being teaching the machine to find where that line should go. We walked through a simple learning model based on naïve Bayes, and in the notebook that implemented the model, we highlighted where we could drop in other algorithms. We ran that notebook on Azure ML and generated a graph to test whether our predictions were in line with what we expected.

Exporting our trained model as a PMML file, we included it in a rules-based decision model. We updated our previous chocolate bar recommendation service using this technique and showed that it was very feasible to build a model combining the best of each type of AI. Finally, we discussed our options as Excel power users for executing these combined models.

This theme of introducing advanced topics to explore further continues in our next and final chapter. That extends several topics we glimpsed...