Introduction of suspected fraud detection on the basis of rules and regulations
The assignment
The client was a German insurance company that had set itself the goal of optimizing its claims management processes. The insurance company had already established one of our solutions for detecting suspected fraud in motor vehicle liability insurance.
They now wanted to extend this functionality to other lines of business such as comprehensive insurance, personal accident insurance, personal liability, household contents and residential buildings. The manual check using criteria catalogs had provided good initial indicators, but the need for an automated solution had increased in order to improve the efficiency and accuracy of fraud detection. The client decided to work with FIDA to introduce a new, rule-based software solution to meet these requirements.
Objective
The aim of the project was to create a software solution that not only enables the automated detection of suspected fraud cases, but can also be fully integrated into the existing claims handling process.
A key aspect of this was to enable the specialist department to independently create and validate sets of rules and transfer them to productive operation. The solution also needed to offer the option of simulating sets of rules in advance using historical data to ensure their effectiveness. It was also important for the software to be audit-proof and meet both technical and functional requirements.
Solution approach
We developed a customized software solution called ARE, which consists of three central components: a web service, a web client and a database.
The web service was designed so that it could be seamlessly integrated into the existing claims handling process. It automatically checks claims and delivers results that can be directly incorporated into the insurance company's processing decisions.
The web client was designed specifically for the needs of the specialist department. It enables employees to develop, adapt and validate sets of rules independently. All the necessary functions are mapped in the web client to ensure complete autonomy for the specialist department. New rules can be designed and existing rules adapted here. The platform also allows rules to be simulated against the productive database in order to check their effectiveness. An integrated approval process based on the dual control principle provides additional security. Once approved, a new set of rules can go live either immediately or at a specified time.
The database serves as a central repository for rules and regulations, test data and simulation results. It has been designed to efficiently support both the storage and processing of large volumes of data. A flexible data model was implemented, which forms the basis for storing case data, attributes, tests, simulations and evaluations.
The introduction of ARE at the insurance company was a complete success. The solution made it possible to automate the detection of suspected fraud in the comprehensive, personal accident, personal liability, household and residential building insurance lines. Seamless integration into the existing processes has significantly increased the efficiency of claims processing, while enabling the specialist department to react flexibly and independently to new challenges.
A look at the details: What was important?
A key focus of the project was the design of a user-friendly and powerful platform that meets both technical and functional requirements. The web client is at the heart of the solution, as it offers the specialist department a wide range of options for reacting flexibly to new fraud patterns. The intuitive user interface enables specialist staff to create and adapt rules and regulations without in-depth technical knowledge. This reduces dependency on the IT department and speeds up the implementation of new requirements.
Another key aspect was the simulation of rules and regulations. To ensure that new or adapted rules are effective, they were simulated against historical data sets before being published. Productive data from the last three to five years was used to realistically test the effects of the new rules. The results of these simulations served as the basis for fine-tuning the rules. This iterative approach ensured that the regulations were optimally adapted to the client's specific needs and challenges.
ARE's technical architecture was developed on the basis of a three-layer structure that ensures a clear separation of presentation, logic and data management. This not only increases the maintainability of the software, but also enables flexible further development. The front end was implemented using modern technologies such as Spring, JSF and Primefaces, while Apache CXF and SOAP were used for the web service. The database was implemented on Oracle, whereby care was taken to adhere to vendor-independent SQL standards wherever possible.
Security also played a central role in the project. A specially developed authorization model makes it possible to control access rights granularly and adapt them to the client's requirements. Communication between the web client and web service takes place exclusively via HTTPS, and all user activities are logged in an audit-proof manner. This includes both login behavior and the use of specific functions within the platform.
