Graph-Powered Analytics and Machine Learning with TigerGraph

Chapter 10. Graph-Powered Machine Learning Methods

After completing this chapter, you should be able to:

• List three basic ways that graph data and analytics can improve machine learning

• Point out which graph algorithms have proved valuable for unsupervised learning

• Extract graph features to enrich your training data for supervised machine learning

• Describe how neural networks have been extended to learn on graphs

• Provide use cases and examples to illustrate graph-powered machine learning

• Choose which types of graph-powered machine learning are right for you

We now begin the third theme of our book: Learn. That is, we’re going to get serious about the core of machine learning: model training. Figure 10-1 shows the stages of a simple machine learning pipeline. In Part 1 of this book, we explored the Connect theme, which fits the first two stages of the pipeline: data acquisition and data preparation. Graph databases make it easy to pull data from multiple sources into one connected database and to perform entity resolution.

Figure 10-1. Machine learning pipeline

In this chapter, we’ll show how graphs enhance the central stages in the pipeline: feature extraction and all-important model training. Features are simply the characteristics or properties of your data entities, like the age of a person or the color of a sweater. Graphs offer a whole new realm of features ...