In this article we continue our interview with predictive analytics pioneer Colleen McCue. Part 1 appeared in our prior post.
5) Colleen, you wrote an article on advanced analytics for predictive policing in 2009. Tell us about that, and how have things evolved since then.
That is one of my favorite papers, and grew out of a talk that I gave to the Virginia Commonwealth University Analytics Initiative. What we were describing was essentially “data science” as applied to the operational public safety community before the term was coined.
One of the core concepts of data science involves having the creativity required to leverage foundation-level concepts in support of novel solutions to the really hard, or emerging problems. The reference to Walmart goes back to their extraordinarily effective use of data to inform operations. For example, “risk-based” deployment actually is nothing more than “just-in-time” policing. Similarly, the effective segmentation of the market into meaningful subsets is comparable to the behavioral schema outlined in the FBI crime classification manual; concepts that we have since translated to better understand extremist groups in Africa.
6) Colleen, from your own perspective, how has the currently immense hype surrounding big data now changed the use of or awareness of predictive analytics, particularly in the public safety area?
Big data have been around for a very long time. In my opinion, the more meaningful “big data” evolution has come from developments in data management, processing, and tradecraft. I can recall an industry challenge a few years ago that called for disruptive or otherwise innovative translations of capabilities in other domains to the really hard problems facing the intelligence community. At the time, there were several efforts directed at exploiting capabilities designed to address “big data” problems for Wall Street and hedge funds. Our team stepped way outside the box and looked to high-energy physics and astrophysics—other disciplines dealing with large amounts of data and weak signal detection problems. We did not make the cut, although in retrospect, many of the “promising” financial services algorithms that were performing so brilliantly in 2007 did not realize their promise when the recession hit.
7) Colleen, tell us about DigitalGlobe Analytics —what is your role there? How is your work there connected to the national security and law enforcement applications that you have been involved with during your career?
I am a Senior Director of Social Science and Quantitative Methods at DigitalGlobe, which really embodies the work that I have been doing for the previous 20 years: characterizing behavior in support of information-based approaches to anticipation and influence. The broad functional purview and access to truly amazing data and technology have really enabled us to take foundation level concepts that we developed in the analysis of crime in the United States, and translate them to other problem sets to include analysis of foreign extremist groups and provision of support to humanitarian missions globally. Not only is this work directly informing operations, but it also is enriching our understanding of behavior, particularly violent or other predatory behavior, that will increase our capacity to readily respond to and adapt existing methodologies to new threats in a meaningful way.
Generally, the DigitalGlobe purpose, “Seeing a Better World,” effectively operationalizes the use of data science in the applied public safety and national security domain. “By giving our customers the power to see the Earth clearly and in new ways, we enable them to make our world a better place.” In my experience, the enhanced vision referenced in our purpose includes the insight derived from our work in operational security analytics, which facilitates anticipation and influence in support of meaningful responses and changed outcomes.
8) Colleen, you were featured in the book "Journeys to Data Mining: Experiences from 15 Renowned Researchers." Describe your experience with that tremendous recognition.
It was a real honor to be included in that book, and I owe a tremendous debt of gratitude to my mentor, colleague, and friend, Dr. John Elder, who reached out to the editor of the text and encouraged him to include me. It also was a validation of the work that I had been doing in this space. Due to the operational requirements and constraints in the public safety and national security environment, I often feel like an outlier in the data science field. We rarely have the time or resources to create the truly elegant solutions that you see in other domains; frequently resorting to the analytic equivalent of duct tape and bailing wire as we try to find answers to some of the truly “wicked” problems. Being included in the book was really encouraging and a great win for our community.
9) Colleen, do you have any final words, suggestions, lessons learned, or other advice to data analytics practitioners?
Let the problem guide the solution. There are a lot of analytic “technicians” who have acquired tremendous expertise with a specific tool, technology, or related tradecraft, but the “master carpenters” tend to be rare. There is a move within our community to emphasize the latter in both training and practice. In my opinion, this is the embodiment of data science as a concept and represents the real future of our field.
We are grateful to Colleen McCue for taking the time to answer these questions and to provide us with her insights into the ever-growing field of predictive analytics. Analytics truly is improving our lives, and we anticipate seeing many more such contributions in this area as the major technology developers, particularly MapR, enhance and improve big data innovation and analytics capabilities.