Data Warehousing in the Age of Artificial Intelligence

Chapter 7. Combining Real Time with Machine Learning

Machine learning (ML) encompasses a broad class of techniques used for many purposes, and in general no two ML applications will look identical to each other. This is especially true for real-time applications, for which the application is shaped not only by the goal of the data analysis, but also by the time constraints that come with operating in a real-time window. Before delving into specific applications, it is important to understand the types of ML techniques and what they have to offer.

Real-Time ML Scenarios

We can divide real-time ML applications into two types, continuous or categorical, and the training can be supervised or unsupervised. This section provides some background on the various categories of applications.

Supervised and Unsupervised

Supervised and unsupervised ML are both useful techniques that you can use to solve ML problems (sometimes in combination with each other). You use supervised learning to train a model with labeled data, whereas you use unsupervised ML to train a model with unlabeled data. A scenario for which you could use supervised learning would be to train an ML model to predict the probability that a user would interact with a given advertisement on a web page. You would train the model against labeled historical data which would contain a feature vector containing information about the user, the ad, and the context, as well as whether the user clicked the ad. You would then train ...