Graph Database Tutorial

Certain analytics problems rely heavily on the relationships between people or things which is not effectively supported by relational databases like Oracle and SQL Server. Social media sites like LinkedIn and Facebook need to keep track of multi-level relationships between hundreds of millions of people to support communities. Financial services organizations like CapitalOne and Visa need to track relationships between large numbers of people, money flows and behaviors to detect potential fraud. Enterprise are also form networks held together by relationships. Retailers want to recommend products or services. Project managers want to plan projects which consist of a network of tasks. There are many additional challenges associated with networks or relationships. The answer to the database part of this challenge is the Graph Database.

What is a Graph Database?

Graph Database Management Systems are a type of DataBase Management System (DBMS) based on nodes, edes and properties which are the key concepts of graph data representation. In the graph model, nodes represent entities and edges represent relationships between entities. Graph DBMS excel when solving problems associated with data heavily structured by relationships such as fraud analytics and social media influence analysis. Graph DBMS enables use of graph algorithms executed in a more efficient manner than relational databases like Oracle and SQL Server

These concepts and terms are fundamental to graph theory:

• Graph: a model that represents relationships.
• Node: the entities in a graph that are related.
• Arc, Edge or Tie: a relationship between nodes.
• Directed Graph: a graph or diagram that has direction such as the progress of time. The first node in each pair points to the second node.
• Weighted Graph: a graph where each edge has a quantified degree such as cost or affinity.
• Degrees of Separation: a measure of how far apart nodes are in a graph.

Who Supplies Graph DBMS Software?

Multiple vendors provide Graph DBMS Software. DB-Engines publishes a ranked list of graph DBMS. The top five in that list as of June 9, 2020 are:

• Neo4j : score 48.27
• Microsoft Azure Cosmos DB : score 30.80
• ArangoDB : score 5.38
• OrientDB (an SAP Company) : score 4.82
• Virtuoso : score 2.28

Graph Use Case 1 - Social Networks

Both providers and those who analyze social media have agendas that can be addressed by graph concepts and graph databases. Social media providers seek to serve members with information like: potential contacts to add to contacts list; number of people known in common; number of people in network; list of contacts at level 1, 2, or 3; or recommendation - members who like X also like Y. Social media analyzers want to use social media to advance their agendas. They may want to: identify influencers, find trends, or gage customer sentiment. Graph databases and concepts can serve both social media providers and analyzers.

Add these graph database social network terms to your vocabulary:

• Bridge: an indication that party provides the only tie between clusters.
• Influence: the degree that a party has the power to impact the opinions or actions of others.
• Tie Strength: the degree of overall closeness between parties.
• Homophily: the degree of similarity between parties based on characteristics such as: demographics, psychographics or behavior.
• Multiplexity: the number of relationship roles between parties. People who are friends and who volunteer together have a multiplexity of 2.
• Reciprocity: the degree that parties have a two way relationship.
• Propinquity: a measure of the geographic closeness of parties.

Graph Use Case 2 - Risk Analytics

Sometimes it feels like we are being assaulted by bad actors. Companies and individuals seek to address challenges like: credit card fraud, money laundering, identity theft, insurance claims fraud and fraudulent product orders. Graph databases and concepts can help.

Graphs can include information like: people, claims, transactions, products, stores, credit cards, geographic location, postal addresses, email addresses and electronic addresses. Rapidly growing networks where people share the same postall address or electronic device can be a clue exposing fraud rings. Changes in behavior such as using a charge card in a geographic location not used is a red flag that a credit card has been stolen.

Graph Database References and Links

Check out these Graph Database links:

• Graph Databases - What's the Big Deal? : Favio Vazquez, 2019, Towards Data Science
• Graph Databases for Beginners : eBook from Neo4j
• Graph Algorithms: Practical Examples in Apache Spark and Neo4J : Free O'Reilly book
• Graph Databases : Free from Neo4j and O'Reilly