Italian solutions vendor LARUS Business Automation S.r.l. and Fujitsu Limited plan to leverage components of Fujitsu’s Finplex AI Scoring Platform Service EnsemBiz for LARUS’ new Graph XAI (1) service, Galileo XAI, an offering for the financial services industry.
Fujitsu’s Finplex EnsemBiz offers users functionality ideally suited to financial services applications with added capabilities based on Fujitsu Research’s existing “Deep Tensor” machine learning technology for analyzing graph-structure data. LARUS will begin offering its new Galileo XAI solution from August, while Fujitsu will offer related integration services to customers in the Japanese market.
The combination of LARUS’ graph-powered platform with Fujitsu’s Finplex EnsemBiz makes it possible to achieve highly accurate analysis of graph structure data, a task not feasible with conventional technologies.
Fujitsu and LARUS have developed this technology to mitigate potential risks and costs facing financial institutions, delivering a solution that can automatically detect sophisticated and complex illegal transactions, including self-financing (2) through advanced analysis and machine learning.
“The financial services industry is keen to incorporate advanced AI technologies to a range of applications to gain competitive advantage and to help businesses in the digitalization of payments,” said Stefano Gatti, Head of Data and Analytics at Nexi Payments S.p.A. (3), The European PayTech leader.
“This joint offering from Fujitsu and LARUS addresses a core industry objective to combat fraud and enhance compliance capabilities with connected data science and graph-based approaches.”
“To ease people’s concerns about AI and further Fujitsu’s goal to make AI trustworthy, we’ve added explainable graph AI ‘Deep Tensor’ technology to ‘Finplex EnsemBiz’, which offers users the ability to understand and verify the reasoning behind the outputs of a system’s decision-making processes,” said Masaru Yagi, Senior Vice President of Financial & Retail Solution Business Group at Fujitsu Limited.
“The transparency and peace of mind provided by Fujitsu’s explainable AI technology will allow customers, especially those in the financial services industry, to confidently embrace innovations that deliver an edge in business. Fujitsu and LARUS look forward to offering graph AI solutions to other industry verticals.”
“From our initial collaboration with Fujitsu Research to Galileo XAI solution commercialization as part of the Fujitsu Finplex EnsemBiz service, we are delighted to have embarked on this journey with Fujitsu, a company known for innovation and technological excellence,” said Lorenzo Speranzoni, CEO of LARUS. “We look forward to our continued partnership and the expansion of our offering to customers in the Japanese and global markets.”
Accelerating digital transformation in financial services with AI
Since January 2020, LARUS and Fujitsu have collaborated on a number of proof-of-concept trials in the financial services space, including POCs aimed at preventing the illegal use of credit cards (4) and illegal automobile insurance claims (5), obtaining verification results of increased detection rates and decreased false detection rates.
LARUS, a Neo4j, Inc. (6) top premier partner based in Italy with over 15 years of business-critical project expertise, helps companies around the world design large-scale, data-driven systems based on the connected data science. With a graph-based approach to artificial intelligence and by empowering AI technology with related context, LARUS helps organizations enhance the efficiency of their business operations by extracting actionable insights from connected data.
Fujitsu’s Finplex EnsemBiz is a service primarily for financial institutions that streamlines loan screening processes and detects fraudulent claims by showing the basis for AI predictions. Fujitsu has added functions utilizing its proprietary AI technology “Deep Tensor,” which derives new knowledge from graph structure data representing connections between people and goods.
Unlike traditional relational data, graph structure data represents diverse and complex data in the form of a network. Fujitsu’s “Deep Tensor” technology can analyze and utilize relationships between graph structure data automatically and more efficiently, making it possible to detect illegal transactions such as circular trading fraud that has proven difficult with conventional technologies.
Description of the special characteristics of the new service
Galileo XAI is a graph-based platform for Explainable Artificial Intelligence with powerful data analytics and visualization tools. Galileo XAI takes the explainability outputs given by Finplex EnsemBiz and shows them in a user-friendly format for quick consumption, thereby satisfying a core requirement for the adoption of AI systems.
LARUS will offer the Galileo XAI platform powered by Fujitsu’s Finplex EnsemBiz to current and future customers. In addition to the financial services vertical, LARUS plans to roll out new Galileo offerings to other industry verticals including government, education, manufacturing, and pharma.
Fujitsu will continue to offer related integration services to increase the detection rate and reduce false positives of fraudulent transactions at Japanese financial institutions using the new Galileo XAI service. Fujitsu will additionally work to enhance the business templates of Finplex EnsemBiz, such as alternative lending (7) and expand application areas to new use cases to realize advanced data analysis in various global fields.
(1) Graph XAI:
In contrast with the usual “black box” in machine learning, Graph Explainable AI allows humans to understand the reasons behind the machine learning recommendations, ultimately increasing confidence and adoption of services based on machine learning.
A type of fraudulent transaction, even if individual transactions appear normal, when the relationship of the different transactions are analyzed together, circular or loop like patterns may become apparent.
(3) Nexi Payments S.p.A.:
Nexi is the leading PayTech company in Italy, the reference point for the digital payments in the country. The company, listed on the MTA of Borsa Italiana, operates in strong partnership with almost 150 partner banks.
(4) Proof of Concept for Credit Card Misuse:
Data from credit card companies in Italy were used to demonstrate fraud detection in circular transactions. Fraud detection rates have been improved from 72% to 89%, and false positives have been reduced by 63% compared to traditional methods using manually defined rule bases.
(5) Proof of Concept for Improper Automobile Insurance Claims:
Data from an Italian auto insurance company was used to demonstrate detection of fraudulent insurance claims. Using a manually defined rule-based approach, the researchers improved detection rates from 18% to 81% and reduced false positives from 82% to 19%.
(6) Neo4j, Inc.:
The developer of the most popular open source graph database system.
(7) alternative lending:
Alternative lending is a way to provide loans using the data, such as trading data and settlement data, which has been used in conventional loans. It is intended to provide loans to people who are unable to access or cannot derive favorable terms from conventional loans provided by financial institutions.