Fraud analytics expert to speak at Discovery Summit 2012
“Analytics are no longer just a nice thing for an organization to have,” says Antonia de Medinaceli, Director of Fraud Analytics at Elder Research, a leading consultancy in data mining, predictive analytics and text mining. She believes analytics are a must-have.
We've invited her to speak at Discovery Summit 2012 on Sept. 12 at SAS world headquarters. She'll explain that while predictive analytics used to give organizations a leg up on the competition, now its use is the norm.
At last year’s Predictive Analytics World in New York, SAS Editor Anna Brown interviewed de Medinaceli about data mining trends and how to build a business case for the use of analytics in your organization. Here is the abridged version.
Have you found any surprising trends in your data mining work?
One of the biggest surprises, even after this many years, is the impact of external factors on a data mining project. As data miners, we're interested in getting to deal with the data, starting to build models – that’s the very exciting part for data miners. But most of the time, we have to deal with political issues, institutional issues, whether it's access to the data or data cleanliness or political situations within an organization. Acceptance of data mining by upper management can be a hurdle that needs to be overcome. So we have all these peripheral issues that as consultants we have to deal with before we can get to the very exciting part of actually doing the technical work.
What are some new approaches you see coming about in the next couple of years in the field of data mining?
There’s really a trend away from developing new algorithms and trying to perfect the algorithms that are already existing. That seems to be pretty established, at least in the traditional data mining side. On the text mining side, however, there's quite a bit of research being done. Industry standards and best practices are still being established. And people are still feeling their way in how text, unstructured data, can help augment the accuracy of a lot of the data mining models out there. There's also quite a bit of work being done in the fraud detection arena, which is certainly my area of interest.
What are some best practices for helping to build the business case for predictive analytics?
What we typically recommend to our clients is to find a champion within the organization or a pain point where predictive analytics can really be shown to be a very valuable technique to use within the organization. And we often recommend doing a quick win, a quick proof-of-concept project before engaging in a long-term analytical project. [We recommend] doing something fairly quickly with a small data set – we call it “hitting a single” – and then you kind of go for the home run later.