Last month I attended the Predictive Analytics World conference in Washington DC. It attracted more attendees from last year with broad representation from multiple industries.

The conference was kicked off by Eric Siegel’s keynote focusing on Five ways predictive analytics cuts enterprise risk. The gist of his presentation was to look at predictive analytics as a data-driven way of exposing risks at a micro-level across different domains within an organization. For example, insurance companies should expose risks in rate making, insurance fraud, customer retention, customer credit in order to influence and impact risk-based decision making at a macro-level. As always, a call out for text mining and text analytics was in store from several presenters. John Elder’s session highlighted text mining related case studies in Federal Government agencies like DoD, SSA, and DHS. A latest survey conducted by Rexer Analytics also revealed that text mining is currently used by a third of data miners in their analyses. I think this is somewhat on higher end but what do you think?

The SAS Platinum Sponsor presentation slot was delivered by Anne Milley. The topic was, "Analytics: the beauty of diversity." Anne’s talk highlighted the need for bringing multiple paradigms (including business users vs. power user; supervised vs. unsupervised learning; visual/interactive GUI vs. programming; analytic application vs. generic tool) to ensure that customer’s needs for discovery, model development and model deployment are met in efficient and effective manner. Anne also showed integration between R models and JMP Pro for users to display analytic results as JMP’s interactive graphics and deliver R programs to a broader audience.

The 2-day conference also featured several interesting customer case studies where predictive analytics was deployed to make variety of decisions. For example:

• Monster.com gave an overview on how to grow business analytics usage across all regions to increase customer retention, market share and customer profitability.
• SunTrust Bank identified situations where using multiple modeling techniques is beneficial and by combining models to get one output for precise targeting of retail banking clients.
• CIBC ‘s Daymond Ling focused on age-old customer segmentation topic and how to leverage it for increasing marketing effectiveness.
• Macy’s Paul Coleman delivered a passionate talk on flat files versus relational database storage approaches for analytic data preparation and predictive modeling.
• UPMC showed its ensemble of boosted decision trees to predict who is likely to be readmitted in the hospital with greater accuracy.

It was a pure coincidence that World Statistics Day 2010 fell on Oct. 20, the second day of the conference, but we were ready to celebrate. Since the conference was in Washington DC, attendees from federal agencies appreciated the day for bringing statistics in forefront. On other hand, many attendees were simply not aware that the United Nations had designated that day World Statistics Day. Statistics is behind a lot of the decisions and policies that organizations (public and private) and countries (developed, developing or underdeveloped) take to achieve better outcomes, improve lives or simply learn from the data they collect.

On the same day, speaking of coincidence, my 8-year-old daughter told me that they were going to learn about mean, median, mode, etc. in her math class using the data students had collected on how long their paper planes flew on playground. It was satisfying to learn that the school system is taking a practical approach to teaching statistics and generate interest. We just need to keep up the excitement for statistics and its applications at all levels – individuals, companies, organizations, and government.

Predictive Analytics World is continuing its journey overseas (UK) and on West coast (San Francisco) in few weeks and months. I hope you can attend the conference next time around!