Graph databases are everywhere, and will become even more ubiquitous

TigerGraph: Unlocks value from connected data with ‘explainable AI’

Paid Feature Graph theory, a branch of mathematics that dates back to the 18th century, is widely applied today in business applications and services. For example, the recommendations you get when visiting social networking and ecommerce sites are the result of harnessing graph databases and analytics to understand entities such as people, places, things, events and locations, and the relationships among them.

Gartner predicts that graph technologies will be used in 80 per cent of data and analytics innovations by 2025, up from 10 per cent in 2021. “A growth in adoption from 10 per cent to 80 per cent over a four-year period is just massive,” says Todd Blaschka, COO at TigerGraph, the maker of the eponymous graph database. “What graphs provide is the ability to identify new patterns and new similarities within data and to do predictive modelling. They help to create better applications for healthcare, manufacturing, supply chain, financial services and many other industries.”

Today’s analytics models require a greater variety of techniques to understand the relationships in the data, incorporate real-time data, and increase the context in AI/ML models. Hence, TigerGraph has created a scalable graph database and analytics platform that unlocks real value from connected data, for technical and non-technical stakeholders alike.

TigerGraph’s ‘Graph For All’ approach has led to low-code or no-code analytics that allows more people to work and interact with data and to glean meaningful insights from complex data relationships. For example, non-technical users can produce and run graph queries simply by drawing the patterns they want, similar to visual data modelling.

Ahead of the curve

TigerGraph recently teamed up with The National University of Singapore Business Analytics Centre (NUS BAC) to make graph analytics capabilities more accessible to enterprises by developing a talent pool with sought-after graph database skills.

The partnership equips students in the BAC’s Master of Science in Business Analytics (MSBA) programme to fulfil the demands of a growing number of businesses turning to graph database and analytics, to facilitate rapid decision making.

Students will also undergo several months of industry internship, during which they work on graph analytics projects relating to critical areas such as fraud detection and supply chain.

“We can potentially create an end-to-end systematic graph analytics training that encompasses the foundation, the theory, the tool and the real-world use cases for NUS MSBA students,” says Associate Professor James Pang Yan, Co-Director of the NUS BAC. “In the future, we can also collaborate with TigerGraph to conduct basic research in the graph analytics areas.”

In 2023, TigerGraph will sponsor the NUS-TigerGraph Innovation Challenge. This challenge, which aims to showcase innovative solutions from NUS students, features problem statements based on realistic business challenges from TigerGraph and its customers.

Growing graph fervour

As graph analytics use cases continue to expand, Blaschka highlights three trends – compliance/ fraud detection, ‘Customer 360’ and digital twin – that have wide impact across many industry sectors.

“In the world of compliance, fraud and money laundering represent costly areas, especially to financial institutions as well as to ecommerce companies,” he says. “For example, trends such as Buy Now, Pay Later (BNPL) and short-term credit services have created a need for businesses to do a real-time credit score analysis and then to match the consumer with the right lender.” The bottom line is that graph analytics enable businesses that offer BNPL to uncover patterns and insights within a deluge of consumer data that can expose risks or fraud.

Another trend is Customer 360’ which aims to achieve a comprehensive understanding of the attributes around a particular customer or user. This trend is evident in the healthcare sector where TigerGraph customers are using graphs to better understand patient needs and expectations, based on data gathered from all patient touchpoints, including their interaction with telemedicine apps.

“By analysing connected patient data, healthcare providers can provide relevant and timely next-care or best-care path recommendations to patients,” says Blaschka.

A third area is creating a “digital twin” or replica of the business application. A twin helps an enterprise find new relationships within historical and real-time data by simulating and modelling what-if scenarios. Automobile makers, for example, have to optimise production resources through supply chain analysis in response to chip supply and shifting market demand for petrol, hybrid or electric vehicles.

“We have customers who refer to graphs as a what-if engine, meaning that you are able to do predictive modelling based on the data and how it changes in real time,” Blaschka explains. “You can apply digital twin modelling to almost every industry sector. Graph is really good at identifying and modelling with fresh data so you can make real decisions on real data in real time to gain a competitive advantage.”

Showing the AI maths

Beyond use cases, graph analytics also addresses a major concern hindering AI adoption – the lack of transparency as to how AI systems arrive at a particular decision. Neural and deep learning networks, for instance, often cannot elucidate the key variables that lead to a decision.

“Graph allows you to explain your AI,” says Blaschka. “Graph can show where the data is coming from, what kind of computation is being done on the data, what is the actual algorithm, what were the weights put on particular parts of the model and then how the results are derived. We refer to this as Explainable AI.”

The use of graph databases augments existing tools to match the needs of advanced data analytics and business intelligence (BI) in the enterprise. In data processing or the data pipeline, data can come from a variety of sources, including existing analytics engines.

For example, enterprises can run graph algorithms such as the Louvain Modularity Algorithm for community detection to understand the relationship of entities within a big community. Then, results from these algorithms can be fed into existing BI tools or other systems.

“The integration into BI tools or through other UIs allow our customers to develop their specific use case,” says Blaschka. “Businesses can now get more insights than they could before to solve a wider range of complex business problems.”

Many of TigerGraph’s innovations are open source. As part of its global support for a thriving developer relations community, TigerGraph runs a series of Graph + AI Summit events worldwide focused on enabling companies, individuals and practitioners to promote the use of graph algorithms in analytics, AI and machine learning projects.

Sponsored by TigerGraph.