In order to illustrate how to use graph to catch fraudsters, we are going to use an example put together by Kenny Bastani on GraphGist. If you haven’t visited GraphGist yet, I suggest you do it : it’s a great resource for graph-related use cases.
When thinking about a graph project, the first step is to come up with a data model. The data model is the way we are going to represent the data we have. In order to build a good (graph) data model, it is usually a good thing to think about what we are trying to accomplish.
Here, we want to detect fake identities. To do that, we are going to focus on the pieces of information banks have on their customers. We want to look for shared information. Fraudsters tend to recycle pieces of information, hence finding shared information will put us on the trail of fraudsters.
The graph model we are going to use will be centered around the customers and the information they provide the bank. Here is a picture of a graph model for fraud detection :