Book Image

Hands-On Graph Analytics with Neo4j

By : Estelle Scifo
Book Image

Hands-On Graph Analytics with Neo4j

By: Estelle Scifo

Overview of this book

Neo4j is a graph database that includes plugins to run complex graph algorithms. The book starts with an introduction to the basics of graph analytics, the Cypher query language, and graph architecture components, and helps you to understand why enterprises have started to adopt graph analytics within their organizations. You’ll find out how to implement Neo4j algorithms and techniques and explore various graph analytics methods to reveal complex relationships in your data. You’ll be able to implement graph analytics catering to different domains such as fraud detection, graph-based search, recommendation systems, social networking, and data management. You’ll also learn how to store data in graph databases and extract valuable insights from it. As you become well-versed with the techniques, you’ll discover graph machine learning in order to address simple to complex challenges using Neo4j. You will also understand how to use graph data in a machine learning model in order to make predictions based on your data. Finally, you’ll get to grips with structuring a web application for production using Neo4j. By the end of this book, you’ll not only be able to harness the power of graphs to handle a broad range of problem areas, but you’ll also have learned how to use Neo4j efficiently to identify complex relationships in your data.
Table of Contents (18 chapters)
1
Section 1: Graph Modeling with Neo4j
5
Section 2: Graph Algorithms
10
Section 3: Machine Learning on Graphs
14
Section 4: Neo4j for Production

Building a link prediction model using an ROC curve

In all the graph analytics problems we have studied so far, our observations were the nodes of the graph. Now, however, we are moving on to a different concept where the observations are the edges. Each row of the dataset should contain information about one edge of the graph. Since our goal is to predict whether a link will appear in the future or is missing from our current knowledge, we can turn the problem into a binary classification one, that is, the edge can either have:

  • the class True, the link exists or is likely to be created, or
  • the class False, the link is very unlikely to appear.

Since we are about to build a classification model, our dataset must include both existing and non-existing edges (the two classes of the binary classifier).

Importing the data into Neo4j

The data we are going to use in the rest of this chapter is a randomly generated geometric graph. This kind of graph has many interesting features, one of them...