Mastering Azure Analytics, 1st Edition

Chapter 9. Intelligence and Machine Learning

Advanced analytics pipelines often include machine learning to make predictions against data flowing through the data pipeline. Azure provides a few ways you can integrate machine learning into your data pipeline. While this chapter is not meant to provide deep coverage of machine learning in Azure, which merits a book of its own, it will provide broad coverage of the machine learning options available and the two most critical phases of most machine learning: model training and model operationalization (see Figure 9-1).

In the world of machine learning algorithms, there are two broad categories of algorithms (aka learners) that are defined based on the way the model is created. Supervised learners are like students in school—they need to be taught by example. They are shown lots of examples and the resulting outcomes, with the goal that one day (e.g., during final exams) they can be presented with new data and accurately predict the outcome (e.g., pass the test). They are called “supervised” because they need training before they can make any predictions, much like students need to be taught the subject before they hope to pass an exam on it. In machine learning, examples of scenarios that use supervised learners include:

• Examining email to classify it as spam or not spam

• Reviewing consumer profiles to predict their likelihood of default on a mortgage

• Predicting flight delays given flight and weather historical data

Figure ...