About Link Analysis

Link Analysis is a data mining technique used to find networks of related entities that appear to be exhibiting some behavior of interest to the business. For example, a network can be a set of brokerage accounts linked together by some chain of internal funds transfers, a set of corporations linked together by wire transfers among themselves, a set of people related as beneficiaries in a set of wire transfers, or a set of people who share a common address, deposit, or name.

It is useful for detecting money laundering and fraud because criminals often work in fraud rings. Fraud rings often contain a series of linked entities that are not normally linked together for business. Link Analysis enables you to detect these networks.

As with any data mining technique, you follow these steps:

1. Identify the behavior in which you are interested. Link Analysis is an appropriate data mining technique if the behavior of interest is based upon a group of related entities.
2. Determine the category of possible related entities and the manner in which they are related. For example, various categories of accounts may exchange money through wire transfers.
3. Identify the characteristics of this group, or network, that are interesting or indicative of the behavior.
4. Determine the patterns that distinguish networks that are exhibiting the behavior of interest, from those that are not.

This thought process prepares you to use the Scenario Manager to create Network Definitions and Patterns that detect the behavior of interest in your data.