A new post on the ZyLAB blog has the title Automatic fraud triangle analytics made possible with text-mining and content analytics. The detection of fraud involves a number of things, but two things in particular arise here. One is an understanding of what drives fraud and the other is the ability to analyse data sources comprehensively.

The “fraud triangle” derives from the work of a criminologist called Dr Donald Cressey who identifies three elements common to nearly all fraudulent activity: the fraudster must have some incentive or pressure (the need for money is an obvious one); there must be opportunity (he or she is in a trusted position with access to bank accounts, passwords etc); and there must be what Cressey calls rationalisation (which might involve, for example, a sense of injustice at being denied promotion or something similar).

The ability to “break” the fraud triangle is obviously helpful in identifying that fraudulent activity has taken place; it is even more useful if a company can anticipate such activities by determining that the three components of the fraud triangle exist.

ZyLAB’s article observes that it is not enough to look merely at the purely financial documents held by a company. The pointers to the possible existence of incentive, opportunity and rationalisation may lie in emails, wordprocessing documents and all the other data which exists in unstructured form in companies’ data stores. This may involve looking for lexical, syntactic and semantic patterns, and delivering timely and comprehensible reports. This sort of pre-emptive search must have regard to confidentiality, privacy and data protection concerns for its use to be considered acceptable.

ZyLAB’s blog article refers to a white paper on the use of text-mining for fraud triangle analysis. You can find that here