Creating and Using Automated Profile Matching Technology

Many financial organisations (i.e. banks and insurance companies) have databases on individuals and their major transactions. These databases have different uses such as obtaining a historical record of an individual. Often the (legal) circumstances mean that a unique identifier for each individual does not exist or cannot be used. Instead personal identification details (names, addresses, dates of birth etc.) must be used to “index” and retrieve records. This is an inexact method and often results in ambiguous situations where multiple records are returned for an enquiry. Disambiguating these situations we term the profile resolution problem. Organisations usually use human experts (review clerks) to resolve ambiguous cases. This is time consuming, expensive and does not address the cause of the problem. We have shown that it is possible to automate the decision making expertise of these review clerks with a number of techniques from artificial intelligence and statistics. We illustrate that the profile resolution problem is actually one of three interrelated problems associated with ambiguous cases and discuss how automated techniques can address each.