Graph Analysis
Graphical structures can be used to model complex domains. It’s possible to leverage graph analysis techniques to yield statistics about a domain model that are not necessarily self-evident. From the explicit connectivity defined by edges (i.e. links, relationships), there exists higher-order patterns that are exposed by understanding the implicit paths through the model.
Graph analysis is all about taking a complicated structure and reducing it to a few understandable numbers. It draws on both graph theory (which is a branch of mathematics that dates itself to the 18th century) and network science (which leverages advances in computing and algorithms). The graph analysis service can help any organization expose information about their domain and as such, drive aspects of their product, marketing, and research.
The graph analysis service typically follows the pattern outlined below.
- Transform a dataset (that is logically a graph) into a format that can be readily consumed by graph analysis software.
- Make use of graph analysis software to compute classic metrics. Such metrics expose the general connectivity patterns of the graph and help point to areas where further analysis may prove fruitful.
- Focus on the client’s particular data analysis needs or perform an exploratory analysis.
- Who are the most influential members of the graph?
- How have particular information pieces disseminated over time?
- What characteristics of the individuals yield connectivity?
- Which products cluster based on purchasing behavior?
- How does connectivity correlate with the growth of our service?
- Provide the client code, reports, and a presentation of the analysis. While metrics are good, its important to explain the meaning of such metrics and how they can effect decision making within the organization.
Finally, if desired, aspects of this service can also be deployed as an active software system for on-demand analysis (both real-time and offline processing possible depending on the complexity of the statistical algorithms). If the graph analysis service is of interest to your organization, please feel free to contact me.
