The Language of Graph is a series of posts that attempts to tame some complexity, clarify graph terminology, and help business leaders understand the power that graph technology can bring to their business.
The previous two posts in the series explained the basics of Graph Query and Graph Analytics. Graph mining describes the process of identifying, quantifying, and cataloguing patterns, associations, and correlations in a large graph. It does this by tracing identifying clusters or substructures in an underlying graph. At its core, graph mining finds similarities among items and groups them.
Uncovering patterns in data allows high-level grouping of elements, whether that be audience segmentation, network anomaly detection, or contact classification. The business value of such grouping can be enormous. For example, building a better understanding of a target consumer market’s interests helps a company to better focus their product development.
What Is Graph Mining?
A simplified definition of graph mining is specialized pattern detection algorithms. In practice, we seek to predict how the structure and properties of a given graph might shed light on some real-world application. Mining also commonly involves developing models that can generate realistic new subgraphs that match the patterns found in large graphs of business data.
What Are the Graph Mining Discovery Patterns?
The most used discovery patterns for graph mining include:
- Graph Pattern Mining identifies frequent subgraphs in one or more sets of graphs.
- Graph Classification uncovers and characterizes the difference between graphs to make new patterns easier to find.
- Graph Compression finds patterns that represent regularities in the data. These data frequencies are used to compress or summarize data better.
How Is Graph Mining Used?
Typical uses of graph mining include:
- Graph Model, Laws, and Generators: synthetic graphs for simulation studies (random graph models, preferential attachment models, optimization-based models, etc.).
- Graph Dynamics Social Network Analysis: the process of investigating social structures through networks and graph theory.
- Graph Summarization and Graph Visualization: a method of condensing and simplifying such datasets.
- Graph Clustering: the clustering of data in the form of graphs. Vertex clustering seeks to cluster the graph's nodes into groups of densely connected regions based on either edge weights or edge distances.
- Link Analysis: the identification of relationships among nodes through visualization methods such as network charts or association matrix.
What Is Graph Mining's Function in Business?
Graph mining seeks meaningful patterns in existing knowledge graphs that have the potential to make changes in a company's operation or improve business processes. Businesses can apply graph mining methods to realize opportunities regarding product innovation and improved services or to transform research and development.
A future post in The Language of Graph series will reveal the meaning of Graph Artificial Intelligence and discuss its role in graph technology and high-performance computing.
For a deeper understanding of the Katana Graph Intelligence Platform, see this three-part series:
Part 1: What is the Katana Graph Intelligence Platform?
Part 2: Katana Graph Intelligence Platform: Data Flow and Strategy
Part 3: Katana Graph Engine
Let’s get started today so that we can help you with your graph needs.
