Roundtable gathers in San Francisco

San Francisco

The fall 2014 meeting of the INFORMS Roundtable was held at the Hilton San Francisco on Nov. 8-9 in conjunction with the INFORMS Annual Meeting. The conference theme was “Real-Time Analytics.” 

The meeting began with a networking trip to Alcatraz Island on Saturday afternoon. The trip included a boat ride to the island and a guided tour of the former federal penitentiary. The enjoyable outing and the gorgeous weather provided an ideal environment for interaction between Roundtable representatives and guests.

Following the outing, the main program began on Saturday evening with a talk on event stream processing for power grid analysis by Brad Klenz from SAS, Inc. Brad spoke about their work on analysis of “phasor measurement unit” data from Duke Energy’s grid. Using advanced analytics techniques, they detect events such as momentary line trip, generator trip, etc., in real time and determine the impact of these events on system stability.

The first talk on Sunday by Deepak Turaga from IBM T.J. Watson Research Center continued on the theme of event stream processing systems and their applications. Stream processing is a new paradigm for gathering, processing and analyzing high-volume, heterogeneous, continuous data streams to extract actionable results in real time. Deepak introduced the programming models and implementation principles for stream processing, discussed emerging analytic algorithms for classification, anomaly detection and pattern mining, and introduced several real-world, big data applications spanning various domains.

Roundtable representative Srinivas Bollapragada from GE Research spoke about real-time algorithms for dispatching trains and routing air planes. Currently at most railroads, dispatchers manually develop meet-pass plans to move trains on the rail network. The algorithm that GE developed to efficiently dispatch trains is currently used at the Norfolk Southern railroad, saving the company several hundred million dollars per year. Srinivas also spoke about algorithms to prevent or minimize unplanned maintenance of airplanes to significantly decrease airline schedule disruption costs

Rohit Tandon from the Global Analytics organization at Hewlett Packard spoke about the value that big data analytics is delivering for both HP and its clients. Rohit described how HP uses predictive analytics to forecast failures of complex digital printers before they occur to better plan field engineer site visits for preventive maintenance. He also discussed how HP’s data scientists partnered with retail analytics practitioners to develop data-driven approaches for creating robust customer engagement strategies.

Rainer Dronzek, the Roundtable representative from McDonalds Corporation, described several current and under-development initiatives at his company related to real-time processing of video and other data to support restaurant operations. His talked included video analytics to assess waiting and service times and other process information, customer usage of restaurants, sentiment and fraud-detection.

To realize the value of analytics in the enterprise, proper choice of architectural patterns is necessary in implementing analytic platforms. These patterns include data lakes, warm-cold data stores, data warehouse augmentation and others. Using examples from insurance, telecommunications, banking, retail and public sectors, David Steier from Deloitte Consulting illustrated the tradeoffs involved in making architectural choices to support analytics. 

Following the regular meeting talks, Jim Williams from FICO introduced the “Analytics Maturity Model” that INFORMS had developed to the Roundtable members. The presentation resulted in a spirited discussion that continued into the meeting with the INFORMS Board later in the day. 

The Roundtable business meeting was conducted by Jeff Winters. The meeting concluded with the INFORMS President’s reception for the Roundtable.

- Srinivas Bollapragada