Insurance Fraud Detection - Using Complex Networks to Detect Suspicious Entity Relationships

The insurance industry is undoubtedly one of the main drivers of today’s economy and certainly will be so for the foreseeable future. This is derived from a basic need every person has, and that is to have financial and personal safety. Despite all this, some entities try to induce fraud on their policies or to circumvent some legal mechanics, to gain unlawful benefits and advantages. This being said, insurance fraud constitutes a grave downside to insurance companies as it directly translates to a loss of economical assets as well as the opportunity to establish a precursor to further exploitation of the system in place. In this context, this dissertation proposes a framework to help detect the most common types of insurance fraud and scam. Most of the times, insurance scams are usually detected after they took place, this means the companies are already at a loss when they detect it. This framework, which is based upon complex networks for relationship visualization, takes into consideration the relationships already in place between the different entities of the insurance hub-world, advises the responsible entities for fraud prosecution on suspicious relationships. This way, frauds and scams can be detected early on, thus minimizing the losses associated. This dissertation is being supported by at Holos, S.A using the online insurance management tool, also known as RIFT. This tool gathers data from actual insurance companies, giving the study a higher degree of veracity and applicability in the real world