Data Mining Interview: Eric Siegel
Filed under: data mining interview, Eric Siegel, Predictive Analytics World
Program chair of the Predictive Analytics World in San Fransisco (more details in this post), Eric Siegel is a professor and data miner with several years experience. He kindly accepted to answer some particular questions about him and data mining for Data Mining Research.
Data Mining Research: How would you introduce yourself in a few lines?
Eric Siegel: I’ve been in data mining for 16 years and commercially applying predictive analytics with Prediction Impact since 2003. As a professor at Columbia University, I taught graduate courses in predictive modeling (referred to as “machine learning” at universities), and have continued to lead training seminars in predictive analytics as part of my consulting career.
I’m also the program chair for Predictive Analytics World (www.predictiveanalyticsworld.com), coming to San Francisco Feb 18-19. This is the business-focused event for predictive analytics professionals, managers and commercial practitioners. This conference delivers case studies, expertise and resources in order to strengthen the business impact delivered by predictive analytics.
DMR: Data mining, machine learning, knowledge discovery in databases, pattern recognition, etc. Are these fields really different?
ES: These overlap greatly, but the terms differ in how specific a method they entail. Saying you’re “mining” through data to discovery useful knowledge doesn’t narrow down the realm of techniques; “data mining” and “knowledge discovery” don’t necessarily refer to any particular methods except to imply one is undertaking a “well-designed” or “advanced” one. On the other hand, machine learning, a.k.a, supervised learning, usually refers specifically to methods such as decision trees, neural networks and logistic regression, which automatically discover predictive models.
In the non-academic commercial world, the term for machine learning is predictive modeling, and, in some contexts, “data mining” refers specifically to predictive modeling. The predictive models derived are ways to describe recurring patterns, so the term “pattern recognition” applies as well.
DMR: What is the most common data mining question you have heard?
ES: It’s a tie. First: “Do you have a case study to clarify the business benefit of commercially deploying predictive analytics and to prove its success?”
Answer: Yes! In fact, the program for Predictive Analytics World is designed for this very purpose, consisting primarily of a veritable warehouse of named predictive analytics case studies across verticals and across business applications. Check out the agenda here.
And second: “Is there risk in deploying a predictive model, relying on its predictive scores to drive operational decisions?”
Answer: Any such risk can be managed as tightly as required by deploying your predictive model incrementally. Once you have a predictive model ready for deployment, start by deploying it in a “small dose”. Keep the current, existing method of decision-making in place, and then – perhaps 5% of the time – employ the predictive model. This way, it stands in contrast to how decisions are made currently, so you can see whether indeed the value of the model is proven – that profits have increased or that response rates have increased.
DMR: Imagine that I can give you any data set by tomorrow. What kind of data would you like mining?
ES: The most predictive data is behavioral data rather than demographic – a person’s (customer’s, employee’s, applicant’s, etc.) behavior is best predicted by their prior behavior – what they’ve *done* rather than who they *are* (a meaningful case against stereotyping, in fact). Give me the transactional history, the online behavior, and the calls to customer service.
This is my wish list, right? So I’ll keep going. The data is big, both wide and long. It is wide since there are many behavioral attributes for each individual or customer. And it is long because we have data for many individuals. In my fantasy, we have half a million rows. But, for the record, a few thousand is often enough.
Finally, the data pertains to a prediction goal for which there is a viable deployment scenario that delivers a strong impact for the business. For example, if we’re predicting customer defection, there’s a lot to be gained for each customer retained, and even more to be gained by targeting retention efforts towards customers predicted to leave.
DMR: What is Predictive Analytics World and who should attend to this event?
ES: Predictive Analytics World, Feb 18-19, 2009 in San Francisco is the business-focused conference that covers today’s commercial deployment of predictive analytics, across industries and across software vendors. In a nutshell, PAW is a warehouse of case studies.
And the leading enterprises have responded, signing up to tell their stories. PAW-09 will have 25 sessions across two tracks, so you can witness how predictive analytics is applied at 3M, Acxiom, Affiliated Computer Services, Charles Schwab, Click Forensics, Google, Linden Lab (Second Life), The National Rifle Association, Pinnacol Assurance, Reed Elsevier, San Diego Supercomputer Center, Sun, Telenor, Wells Fargo Credit Card Services, Wells Fargo Internet Services Group — plus special examples from Anheuser-Busch, Disney, Hewlett-Packard, HSBC, IRS, Pfizer, Social Security Administration and WestWind Foundation.
For a summary of business applications of predictive analytics – and a named case study for each – see my article, “Predictive Analytics Delivers Value Across Business Applications” here
The number one Netflix Prize competitor, who recently won the Netflix Progress Prize, will reveal their secret sauce, and you’ll hear from several industry thought leaders, including keynotes from Yahoo!’s Chief Data Officer & Executive VP, and Amazon.com’s Former Chief Scientist. The conference kicks off on a hot topic with my keynote, “Five Ways to Lower Costs with Predictive Analytics”, and ends with two predictive analytics workshops that serve as a third-day option.
With such a range of speakers and case studies, I’m super-excited about this program – there’s nothing else like it!
The conference program is designed to speak the language of marketing and business professionals using or planning to use predictive analytics to solve business challenges. Since the best way to catalyze commercial deployment is to show the people it really works outside “the lab”, PAW’s program is packed primarily with named case studies of commercial deployment. And for the hands-on practitioner or analytical expert focused on commercial deployment who wishes to speak this same language, it’s an equally valuable event.
For informative event updates, sign up here.
You can find another interview with Eric Siegel by Romakanta.