Summary

In this chapter, we have moved away from the Amazon ML web interface and learned how to work with the service through the AWS CLI and the Python SDK. The commands and methods for both types of interaction are very similar. The functions and commands perform a standard set of operations from creation to deletion of Amazon ML objects: datasources, models, evaluation, and batch predictions. The fact that Amazon ML chains the sequence of dependent object creation allows you to create all the objects at once without having to wait for one upstream to finish (datasource or model) before creating the downstream one (model or evaluation). The waiter methods make it possible to wait for all evaluations to be completed before retrieving the results and making the necessary object deletion. 

We showed how scripting Amazon ML allowed us to implement Machine Learning methods such as cross-validation and Recursive Feature Selection, both very useful methods in predictive analytics. Although we...