There is an enormous pool of data, much of it without barriers. The potential to mine data for practical application is now not only possible, but critical as a business driver.
Moreover, Big Data has an important role to play in predictive analysis and its particular purpose in minimising fraud.
“Big Data is the ability to search, aggregate and cross reference different data sets both structured (mobile number) and unstructured (email),” says Manoj Chiba, senior advisory consultant at the Gordon Institute of Business Science.
To assist fraud profiling, predictive analytics looks at two data streams. One is descriptive and is concerned with the characteristics of those who commit fraud. Predictive analytics looks at how likely a claim is to be fraudulent when submitted by an individual or business with those characteristics. This enables the establishment of rules for better decision making once the data is understood.
Predictive analytics enables decision making to move from a base of uncertainty to a base of usable probability, Chiba says, go from knowledge to action.
The following case study illustrates the effect of predictive analytics.
A financial institution was experiencing a large number of credit card fraud incidents.
Typical behaviour or symptoms indicated a small purchase followed by a big one; large number of online purchases in quick succession; spending as much as possible quickly; smaller amounts, spread across a longer time.
The problem was to identify characteristics of transactions that deviated from the norm. Age and income would be factors, even time of expenditure. For example, an early morning purchase appearing as irregular for the user would be an alert (near real time detection of fraudulent purchases).
More than 2 million data points were used, the case involved more than 350 hours of pure analysis and three months to understand the data.
The result for the business was that it could detect fraudulent purchases in credit card use with a 76% to 85% accuracy rate.
Some challenges for business
Chiba identified the key characteristics of successful fraud analytics models as: statistical accuracy; interpretability; operational efficiency; justifiable cost and regulatory compliance.
He says, “Only 34% of global companies recognise the value of data.”
Performed by a data scientist, a recent product of the Harvard Business School, predictive analytics is the branch of the advanced analytics which is used to make predictions about unknown future events. Predictive analytics uses many techniques from data mining, statistics, modelling, machine learning and artificial intelligence to analyse current data to make predictions about the future.
Data scientists are rare. These individuals must apply thinking and expertise from a variety of fields to solve problems. The job demands problem-solving skills – logic and reasoning – the ability to know how non-traditional and traditional data sources can assist business to derive and drive value.
Chiba says predictive analytics is not new but there have been some changes such as the increase in volume and type of data; greater interest in data for insights. Computer power along with point and click access has also contributed to the change.
“We also confront tougher economic conditions and the need for competitive differentiation, business efficiency and return on investment, among other things.
“Big Data and analytics provide powerful tools that may improve an organisation’s fraud detection systems but,” Chiba stresses, “is complementary to traditional expert-bases fraud detection approaches and does not replace them.”
Says Manie van Schalkwyk of the Southern African Fraud Prevention Service, “We are confident that we can utilise predictive analytics in our holistic effort to combat fraud. We find this new dimension very exciting and will work closely with our clients to make data analysis more relevant in our approach to reducing fraud statistics.”