Detect insurance fraud with graphical algorithms

While other forms of data analysis can also detect cases of fraud, these other machine learning algorithms often act as “black boxes” – spitting out predictions, but not always giving analysts the necessary context to that they immediately understand why it is likely that a claim could be fraudulent. (Credit: Olivier Le Moal/Shutterstock.com)

Insurance fraud is a huge problem, not only for insurance companies, but also for law enforcement, banks and other financial institutions. According to some estimates, fraud accounts for as much as 10-20% of insurance losses. The FBI estimates that the total annual cost of insurance fraud exceeds $40 billion, with some other estimates exceeding $80 billion. Consumers end up paying for the fraud through higher premiums, which costs $400 to $700 per year for the average American family.

The vast majority of insurers have dedicated fraud investigation teams in place, yet many insurers have not begun to take advantage of modern technology to detect instances of fraud. As late as 2019, only 1 in 5 insurers planned to implement artificial intelligence for fraud detection in the next two years.

The benefits of big data analytics, artificial intelligence and machine learning are obvious. These tech tools can process a lot more information than human teams can handle, and over time they can even learn to spot suspicious behavior better.

In particular, graphical algorithms are extremely useful for analyzing insurance claims data. A graph algorithm (or simply “graph”) is a data structure composed of vertices (various data points) and edges (relationships between these data points). Charts can be useful in areas such as social media and transportation, helping organizations better understand the relationships and interactions between users and vehicles, respectively. They can also help finance and insurance companies more accurately identify instances of fraud.

Somehow the charts seem tailor-made to detect insurance fraud, which is often perpetrated by loosely connected criminal networks. For example, in states with no-fault auto insurance (which allows policyholders to recover losses from their own insurance company, regardless of who is at fault in a car accident), dishonest attorneys, medical providers, repair shops and others may “cap” charges for legitimate claims. In other cases, crime teams will stage completely fake crashes – with fake drivers, fake passengers, fake pedestrians and fake witnesses. To mask their criminal activity, these fraudsters sometimes switch roles — playing the driver in one scam, a pedestrian who was hit in another, and a witness in yet another.

These fake accidents will mostly result in relatively small claims. And since minor motor vehicle accidents happen multiple times a day in real life, a single incident is unlikely to trigger red flags. But by connecting the dots through graph algorithms that identify “core” players or uncover certain patterns in graphs, insurers can find the hidden relationships between multiple crashes and begin to see patterns that could indicate fraud.

Visualization is one of the main advantages of examining data using a graph algorithm. A graph makes it easy for data analysts to see all the different relationships between different actors, and then drill down deeper into an incident or group of incidents once they notice a suspicious pattern. This visual element can also be useful for reporting suspicious activity to management.

While other forms of data analysis can also detect cases of fraud, these other machine learning algorithms often act as “black boxes” – spitting out predictions, but not always giving analysts the necessary context to that they immediately understand why it is likely that a claim could be fraudulent. The visual component of charts allows analysts to immediately see the relationships between different parties in a scam and also gives them the information they need to effectively escalate the case within their organization.

Additionally, analysts can label the vertices and edges of graphs with metadata, incorporating factors such as age, number of times a person has been involved in an accident, and any other information deemed relevant.

In a recent white paper, one of the leading graph database vendors presented an example scenario showing how powerful graphs can be in helping insurance companies detect fraud. In this scenario, criminal networks comprised of doctors, lawyers, auto body shops, and accident participants collude to stage “paper collisions” that result in soft tissue damage. These types of claims are favored by fraudsters because they are difficult to validate and expensive to process. If ten people stage five fake crashes, the white paper estimates, the fraud ring can generate up to $1.6 million in injuries and auto damage claims.

Julian Shun, lead instructor at the Massachusetts Institute of Technology.  (Credit: Lillie Paquette) Julian Shun, lead instructor at the Massachusetts Institute of Technology. (Credit: Lillie Paquette)

Even the most sophisticated data analytics solution will not be able to completely eliminate insurance fraud. But graph algorithms are a powerful tool that can help analysts spot the relationships that form the basis of insurance fraud.

Julian Shun, senior vocational training instructor at the Massachusetts Institute of Technology (MIT) Graph algorithms and machine learning course, is an Associate Professor of Electrical Engineering and Computer Science at MIT and a Principal Investigator at MIT’s Computer Science and Artificial Intelligence Laboratory. His research focuses on the theory and practice of parallel algorithms and programming, with particular emphasis on the design of algorithms and frameworks for large-scale graph processing and spatial data analysis. He also studies parallel algorithms for text analysis, concurrent data structures and deterministic parallelism methods.

The opinions expressed here are those of the author.

Related:

Sharon D. Cole