Chapter 2: Architectures for Streaming and Real-Time Machine Learning

Streaming architectures are an essential component of solutions for real-time machine learning and streaming analytics. Even if you have a model or other analytics tools that can treat data in real time, update, and respond straight away, this will be of no use if there is no architecture to support your solution.

The first important consideration is making sure that your models and analytics can function on each data point; there needs to be an update function and/or a predict function that can update the solution on each new observation being received by the system.

Another important consideration for real-time and streaming architectures is data ingress: how to make sure that data can be received on an observation per observation basis, rather than the more traditional batch approach with daily database updates, for example.

Besides that, it will be important that you understand how to make different...