Graphics Reference
In-Depth Information
Similarities in Fraud Claims
Insurance fraud is a significant problem. Although it might be perceived
as a victimless crime, insurance fraud can affect innocent people directly
through injury or damage and also increases insurance premiums for
everyone using insurance.
One way to explore potential fraudulent claims is to analyze similarities
between insurance claims. One person may have multiple valid claims after
a series of accidents or after a series of thefts. But a fraudster may attempt
to obfuscate multiple claims by slightly altering personal information (for
example, filing one claim under the name Jeff Benson and another under
the name Geoff Bensen) or using slightly different address (for example, 34
Main Street versus 34a Maine Street).
Figure 9-2
shows an export of some data from a fraud analysis system and
that has been loaded into Cytoscape. Each node represents an insurance
claim, and each edge represents one type of similarity between two claims.
The visual attributes of the edges have been set so that similarities are
color-coded (for example, purple for similar phone numbers, green for
similarly named involved parties, and red for similar vehicle identification
numbers, or VINs). A similarity is represented as dashed for a weak
connection, with a solid line for an exact match.
Notice the multiple different fraudulent components (that is, separate
clusters). At the top left is a component with connections between various
claims that all have similar phone numbers (purple connection). In reality,
this may not be a fraudulent cluster (the phone numbers could be different
extensions from one place of business).
The lower-left component has similar phone numbers but also has quite
a few claims with similar VINs (red links). This may be a set of claims
with people all using the corporate phone number, and vehicles may have
similar VINs because they may be part of a company's fleet of vehicles and
purchased in a single large transaction.
