Why predictive modeling of human behavior demands an end-to-end, low-latency database architecture
Here are some key points from the article in addition to some insights about graph analysis and big data:
- Semantic graphs map relationships among words, concepts and other constructs in the human language allowing for unstructured data to be used in a graph showing important connections.
- Graph analysis is not new. It has been used as a form of data visualization to explore connections and identify patterns and relationship that would otherwise have gone undetected.
- Some vendors have taken their graph capabilities to new levels. For example, Centrifuge Systems allows users to draw the graphs, search the graph space, interact with charts and display important measures about the graph network. Analysts can easily pinpoint portions of the graph that require additional analysis. Hotspots of interesting activity jump out from the graph based on the number of connections and important performance measures.
- While social graphs may be the most popular, this approach is especially useful in detecting fraud networks, cyber data breaches, terrorist activity and more.
- One of the most important points is that graphs can incorporate diverse streams of big data including both structured and unstructured. Imagine the ability to analyze banking wire transfer data in the same graph with unstructured data that includes names, locations, and employers - intelligence that has been discovered through the semantic processing of unstructured data. That's a powerful combination of sources linking data from the open web with transactional information. When done in real-time, this can be used in anti-money laundering, fraud prevention and homeland defense.
- "Data scientists explicitly build semantic graph models as ontologies, taxonomies, thesauri, and topic maps using tools that implement standards such as the W3C-developed Resource Description Framework (RDF)."
While this may be beyond the scope of many NoSQL and Hadoop databases, MarkLogic 7 is embracing triple stores as they continue to innovate on their Enterprise NoSQL approach. No one else has values, triple store data derived from semantic processing and documents with real time indexing and search - The bar for Enterprise NoSQL is about to be raised again.
You can read more about this on Semantic Web: