Mastering Machine Learning with Spark 2.x

Property graphs

Before we move on to introduce GraphX as a graph processing engine, let's look at an extension of graphs as we have seen them before. We have already considered labeled graphs as a convenient way to name vertices and edges. In general, the graph data we will consider in the applications will have more information attached to both vertices and edges, and we need a way to model this additional information within our graph. To this end, we can utilize the notion of property graphs.

From the basic definition of a graph as a pair of vertices and edges, it is not directly possible to attach additional information to the two structures. Historically, one way to circumvent this is to blow up the graph and create more vertices corresponding ...