Graphics Reference
In-Depth Information
Figure 9-2: Different clusters of suspect insurance claims are linked by
similarities.
In the middle of Figure 9-2 , stretching from top to bottom is an interesting
component showing many similarity types and multiple connections
between a pair of claims, including similar people names (green), similar
phonenumbers(purple),andsimilaraddresses(brown).Anumberofnodes
have many connections, and these highly connected nodes are each
somewhat unique. Notice that claim 44 (near the top) is mostly connected
bysimilarlynamedpeople,whereas55(right)ismostlyconnectedbysimilar
phone numbers.
Claim 42 shares three solid connections (two people and one vehicle) with
claim 70, which is likely the same vehicle and occupants (for example, a
spouse), but claim 42 is also suspiciously linked to other claims in a wide
variety of similarities. Claim 42 is a claim worth exploring further.
The first step is to investigate the immediate details available in the graph.
The data was exported from the fraud system, including the item labels of
the similarity match. By selecting the edges around node 42, you can see
the details for each link in the Table panel, as shown in Figure 9-3 . You
will notice similar names (note that the real data has been modified here
to reflect fictional names). Based on this set of matching names (as well as
phone numbers and addresses), a fraud investigator may be inclined to go
back to the various connected claims and open an investigation.
 
 
Search WWH ::




Custom Search