Investigative Data Mining Articles

Data mining as a concept has been around for
quite some time. A brief search of the Internet
will reveal many sites that provide information
about this subject with numerous suppliers of
products and services. Expand the search criteria to include fraud detection and the results will include a host of topics such as expert systems, fuzzy logic, neural networks, genetic algorithms and pattern recognition systems. While the suppliers of many of these systems purport to have developed the "Holy Grail" of fraud detection systems, the reality is often quite different.

Many of these systems were initially developed at universities and other research establishments and were designed to identify patterns in immense quantities of apparently random data. In a similar way, the fraud examiner will analyse apparently random corporate data such as call log records, invoicing transaction files or trading histories to determine if some external agent (the fraudster) is distorting the random nature of the data and leaving a discernible pattern.

Purely statistical data mining techniques do not appear to transfer easily from academia to the real world of fraud detection, although as with all generalisations there are always a few exceptions. There are for instance a number of very sophisticated data mining systems that rely heavily on statistical analysis, neural computing and genetic algorithms which have been successfully implemented in such environments as credit card transactions monitoring, telephone billing systems and stock exchanges. Unfortunately, the cost of implementing these can often be several hundred thousand pounds and putting them generally beyond the budgets and capabilities of most commercial organisations.

In practice, the fraud examiner can still make use of data mining techniques without the need for prohibitively expensive resources. Such techniques can range from the simple use of spreadsheets, to analytical audit software, specialised analytical tools or even neural and fuzzy logic systems. Increasingly, the ability to interrogate data is also built into mainframe accounting systems. The fraud examiner should be aware of the types of tools that are available in order to make an informed decision about which are appropriate for their investigation team or organisation.