The Language of Graph is a series of posts that attempts to tame some complexity and demystify graph terminology.
The first post in this series, The Language of Graph Query, defined graph query as the search for patterns in graphs of data. It also discussed industry applications of graph query, including identification of fraudulent activities, streamlining of supply distribution, and flow of information to ensure patient healthcare. The next most significant term to understanding knowledge graphs is Graph Analytics.
What Is Graph Analytics?
A graph represents relationships among items. Mathematically, a graph is simply a set of vertices and a set of edges between vertices. However, this mathematical definition translates into software systems for graph problems that are both diverse and complex. There are multiple ways to represent graphs and multiple abstractions for defining graph algorithms for a given representation.
(Azad et al., IEET)
Graph Analytics, as opposed to Graph Query and Graph Mining, is a term that refers specifically to the process of analyzing problems that compute global properties of an entire population, touching every part of the graph until some convergence property is reached. A common example of this type of problem is computation of page rank. Graph analytics applies algorithms that help analysts understand the complex relationships and dependencies between entities such as people, companies, chemicals, and diseases. Common real-world concerns addressed by graph analytics include:
- Discovering the centrality or importance of an entity within a connected group.
- The differences and similarities between groups within a larger population.
- Understanding the strength of the connection between two distant entities and graph database entries.
Glossary of Top 15 Graph Analytics Terms explains several of the terms used above in describing graph analytics.
How Is Graph Analytics Used?
Examples of available Katana Graph global analytics routines include:
- Connected Components: Connect groups of potential synthetic accounts across multiple channels.
- Louvain Clustering: Identify hierarchies among communities of possible synthetic accounts.
- Jaccard Similarity: Given a new customer, identify similar customers for potential duplicates or fraudulent accounts.
What Does Graph Analytics Do for Business?
Farshid Sabet, CBO of Katana Graph, summarizes the value of graph analytics:
Graph analytics is used effectively in financial, security, health, life sciences, and many more markets. For example, pharmaceutical companies use graph analytics and graph mining for various use cases, and they have employed teams of scientists to develop knowledge graphs to use data for faster drug discovery or more in-depth analysis.
Algorithms such as Connected Components, Louvain Clustering, and Jaccard Similarity help businesses identify opportunities to create new products, to connect customers to the appropriate products, and to cut costs. Graph analytics’ ability to build context and draw insights from complex data has also been demonstrated in a wide range of business use cases including fraud detection, computer network resource management, drug discovery, manufacturing master-data analysis, and supply chain optimization.
Katana Graph’s all-in-one graph intelligence platform features the game-changing intersection of graph technology and high-performance computing.
For a deeper understanding of the Katana Graph Intelligence Platform, please read this three-part series post:
Let’s get started today so that we can help you with your graph needs.
