We’ve all heard “a picture is worth a thousand words”. But is it really? Upon reflection, this truism fails outside of, say, advertising, some works of fine art, and now, perhaps, in the area of computing. When it comes to representing the kind of data we encounter most in life, a picture is worth quite a lot.

We began the series “Graph embeddings for machine learning, the key to unlocking graph data” with an assumption that graph databases are all around us. Well, if you work with data a lot (you’re a data scientist), you probably take this for granted and moved quickly to the more technical and meatier part of the series. If you work with or manage data scientists, or are thinking about it, it’s worth unpacking the critical role graphs play in our day-to-day life.

In part 1 of 3 of our graph series, we defined a graph as a collection of objects and their related links. Said another way, a graph is a sort of mathematical picture that conveys more information than what numbers can alone. Graphs convey space and relationships. As such, graphs represent real-world systems better than lists of numbers.

That’s because most of our universe, world, and day-to-day lives are interconnected. We provided a list in our September 3, 2019 post. Here are a few more examples of graph data:

  1. Tracking or identifying patterns of disease outbreak

  2. App end-user models ( endless examples here, depending on the app)

  3. Legal document classification and organization

  4. Trends in on-time and late payments by clients (read our case study on that here)

  5. Optimal engagements based on geography and client type

We lose granularity when we shift away from informational models that remove the relationship-dimension of information, which we necessarily did earlier in the history of computer science. But now, as we pointed out before, more graph data is more available and we have better methods for depicting and using non-linear information. Computer science has come full circle.

Reducing much of our world into columns and rows reduces the richness and nuance of the data input. Better machine learning systems intake interconnected, “messier” data in a conversion process we will continue to unpack in coming weeks. Things get interesting (and, for those of us building sophisticated information technology systems, things get competitive) when a machine learning model can process data inputs from the real world.

As machine learning begins to take better account of graph data, we can unlock exciting opportunities to automate the analysis of complex and critical systems. Learn more about how we can help or contact us to set up a call to discuss how we can help you unlock your graph data.