The Use of Analytics for Claim Fraud Detection
Roosevelt C. Mosley, Jr., FCAS, MAAA
Nick Kucera
Pinnacle Actuarial Resources Inc., Bloomington, IL
ABSTRACT
As it has been widely reported in the insurance trade news, fraudulent claims continue to be a significant issue in the insurance industry, costing policyholders billions of dollars. More companies are turning to analytics to help identifying claim fraud. Identifying claim fraud using predictive analytics, however, represents a unique challenge.
1.
Most predictive analytics applications have a complete target variable which can be analyzed. Fraud is unique in that there is generally a lot of fraud that has occurred historically that has not been identified.
Therefore, the definition of the target variable is not complete.
2.
There is a natural assumption that the past will bear some resemblance to the future. In the case of fraud, methods of defrauding insurance companies change quickly and it can make the analysis of a historical database less valuable for identifying future fraud.
3.
In an underlying database of claims that may have been determined to be fraudulent by an insurance company, there are inconsistencies between different claim adjusters regarding which claims are referred for further investigation. These inconsistencies can lead to historical databases that are not complete.
This paper will demonstrate how analytics can be used to help identify fraud and allow an insurer to optimize the resources they have available in combating this fraud. Applications discussed include:
1.
More consistent referral of suspicious claims to claim investigative units
2.
Better identification of suspicious claims, even as techniques used to defraud insurers are changing
3.
Incorporating claim adjuster insight into analytics results to improve the process
As part of this paper, we will demonstrate the application of several approaches to fraud identification:
1.
Clustering
2.
Association analysis
3.
PRIDIT (Principal Component Analysis of RIDIT scores)
INTRODUCTION
In the property and casualty insurance industry, there are many reports that the occurrence of claim fraud is increasing, and it is evident that the focus on claim fraud has been magnified. Many of the estimates of the amount of claim fraud in the industry show this trend, and one only has to follow an insurance news feed to know that the amount of reporting that is related to claim fraud seems to be rising significantly. Regardless of the exact
1
amount of fraud present in property and casualty insurance, all agree that it is a significant amount and thus a concern that needs to be addressed.
Insurance companies have developed effective procedures for identifying, investigating, and deterring fraudulent activity. The combination of experienced claim adjusters and special investigators has produced a process that ensures claim payments being made are fair and legitimate. The experience, insight, and intuition of these claims personnel have saved insurance companies millions of dollars in payments over the years.
However, as good as experienced claim adjusters and special investigators are, the reality is there are not enough of these trained eyes to review every claim, and thus there are some fraudulent claims that slip through the cracks.
As a result, payments are sometimes made that should not be.
Predictive analytics can assist insurance companies in developing a more consistent claim referral process, such that the benefit of the expertise of the best adjusters and investigators is applied to all claims. Predictive analytics can also enhance the work of the claims department by uncovering complexities and nuances in a particular claim that may be missed by even the most experienced claim adjusters.
THE CLAIM FRAUD PROBLEM
The problem of claim fraud by any measure is a large one. This can be seen in
