From fraud detection to healthcare, enterprises embrace graph AI for richer insights, deeper understanding, and smarter decision-making.
The use of graph data is rapidly becoming mainstream in enterprise IT. In its Top 10 Data and Analytics Technology Trends for 2020 report, Gartner notes that “understanding relationships and generating graphs based on the combination of diverse data at scale are the keys to modern analytics.”
According to a Gartner survey, 92% of companies will use graph AI and machine learning (ML) in the next five years. An explosion of academic research has bolstered this trend. Over 28,000 peer-reviewed papers about graph-powered data science have been published in the past decade. Globally, graph data science is viewed as a significant research focus.
Achieving Business Success
Graph data science is a powerful and innovative method. It can calculate the “shape” of a connected context for each piece of data using graph algorithms. As a result, it enables ML predictions that are significantly better and richer. Earlier, big tech companies like Google could handle massive volumes of connected data. Today, these approaches are available to all enterprises.
Graph data science can transform how enterprises make predictions in many diverse scenarios, from fraud detection and tracking customers to patient journeys. It uses the connections between data points to make more accurate and interpretable predictions.
Identifying potential new associations between genes, diseases, drugs, and proteins in a drug discovery context means providing the immediate context necessary to assess a discovery’s relevance and validity. Customer recommendations mean learning from user journeys to make accurate suggestions for future purchases while presenting options relevant to their buying history to boost trust in suggestions.
With ML, organizations could go to the next level by quickly learning generalized, predictive features from data. Several companies are still learning to use connected data in their ML workflows. However, there is less friction to get started and more real-world examples and case studies.
In terms of value, knowledge graphs are an excellent starting point. The concepts they illustrate connect ideas and are a foundation for more advanced approaches, such as graph algorithms or deep learning. The potential payback of graph technology encourages data scientists to recognize that, whether it’s finding patterns, identifying high-value features, or training machine learning models, a lot of their work cannot be accomplished without it.
Also Read: We All Have a Tech Stack, But Are We Using It Right?
Use cases
- The British government also uses graph science data in a post for GOV.UK, Felisia Loukou, and Dr. Matthew Gregory discuss deploying their first ML model. Which automatically recommends content to users on the site based on which page they are on – quite similar to how you might get a new show to watch on your favorite streaming app recommended.
- Predictive Maintenance: Another example is Caterpillar, one of the world’s leading construction equipment manufacturers, which gained recognition for its graph data science innovations. For predictive maintenance, the firm uses natural language processing (NLP) and a knowledge graph. Parsing more than 27 million documents, its IT team identified valuable data.
The team developed an NLP tool to discover previously hidden trends and connections. The classification tool uses the data sections already tagged with terms like “cause” or “complaint” to apply to the rest of the data. Stanford Dependency Parser is accessed via WordNet, a lexicographic dictionary. It then analyses text and graph technology to find patterns and connections, build hierarchies, predict outcomes, and add ontologies. Using the newly connected data, users can conduct meaningful, data science-enhanced searches.
- Enhancing The Detection Of Infections: New York-Presbyterian Hospital is another example of graph data science in action. Using the technology, the analytics team tracks infections and takes strategic measures to contain them. Using graph technology, their developers could connect all dimensions of an event — the what, when, and where.
As a result of this insight, the team created a “time” tree and a “space” tree to model all the rooms patients could be treated in on-site. Although the initial model revealed many relationships, it did not meet project goals. An event entity was included to connect the time and location trees. The analytics team can analyze everything in its facilities using the resulting data model. The system can then proactively identify and contain diseases before they spread.
Businesses Are Now Using Graph Data Science
Gartner’s data industry team predicts that a quarter of global Fortune 1000 companies will possess a graph technologies skills base in three years. Their data and analytics initiatives will benefit from graph technologies. By 2023 and beyond, graph-enabled data science will become essential to business analytics. Organizations must focus on understanding it as a key to unlocking business advantages.
