"Calculating Car Insurance Premiums Using Communication Data" Financial Services Commission Selects Top Teams for D-Testbed

On the afternoon of the 23rd, the Financial Services Commission and the Korea Fintech Support Center held the '2024 D-Testbed Awards Ceremony and Case Presentation' at Front1 in Mapo, Seoul.

The D-Testbed project allows participants to test financial-related ideas that have not yet been released to the market by accessing secured information systems online and utilizing provided financial and non-financial data. It started in 2021 and recruits participants twice a year.

The Financial Services Commission Chairman's Award was given to the FraudNet team, who conducted verification with the idea of "Enhancing analysis and detection of suspected financial fraud transactions using artificial intelligence (AI) technology." They analyzed pseudonymized home banking and firm banking (inter-institution exclusive banking transactions) transfer data from the electronic financial joint network using machine learning and deep learning technologies, then analyzed the characteristics of suspected financial fraud transactions and strengthened detection performance.

The AI FinLab team received the Financial Supervisory Service Governor's Award for verifying the idea of "Generating and analyzing synthetic financial data using generative AI models." They generated synthetic data by analyzing financial transaction data such as fund flows between accounts and received praise for improving abnormal transaction detection functions using this data.

There was also a team that verified a car insurance accident rate prediction model by combining financial and non-financial data. The Crepas Solution team, which received the Fintech Support Center Chairman's Award, analyzed correlations between financial data and non-financial data such as telecommunications data and incorporated customer behavior patterns into the car accident rate prediction model. Through this, they also explored differentiated insurance premium calculation methods.

The Financial Services Commission plans to recruit participants for the D-Testbed project twice a year again next year. Starting next year, participants will be able to bring corporate data they own into the D-Testbed, combine it with corporate credit information provided by the Credit Information Center, and conduct tests. Additionally, a new "AI Track" (tentative name) will be established to provide some high-performance analysis environments.