Editor's Desk

By Michael Hughes

Tasty data analysis

Planning for the future is tough, predicting it even tougher.

After all, if businesses knew exactly what their customers wanted and exactly when they wanted it, industrial engineers would have an easier job designing production and logistics processes.

Well, these days, customers, through social media, often are telling the world not only what they want but where they're going to get it. And through Facebook, Foursquare and the like, people do this over and over again. Facebook alone has hundreds of millions of users worldwide. So what would happen if we could pour all that information into an artificial neural network and figure out what happens next?

Well, that's exactly what Keith Thompson, Bichen Zheng, Sang Won Yoon, Sarah Lam and Nathan Gnanasambandam describe in "Understanding Behavioral Patterns," this month's cover story. Starting on Page 28, the authors share a project that used location-based social media to predict restaurant choices.

The team from the Watson Institute for Systems Excellence (WISE) at the State University of New York at Binghamton and the Xerox Research Center used 121,000 restaurant check-ins from the greater New York City area, dividing them into 13 different restaurant types. They considered the shared relationships among users, time of day and historical restaurant decisions. Personal mobility characteristics also were taken into account.

The artificial neural network turned out to be quite cognizant. For each individual tested, the prediction rate was greater than 90 percent.

While great for the restaurant industry that can figure out how to deploy food and workers to prepare the next meal, the work has broader impacts for government and society.

General movement patterns can help planning boards at all levels of government with deciding where to build that new road, bridge or subway terminal. The information also is useful for devising evacuation plans in case of disaster. And such models can help determine how a disease epidemic will spread.

Welcome to the brave new world of big data analysis. Next time you can't decide where to eat, maybe you should just ask the network.

Michael Hughes is managing editor of IIE. Reach him at mhughes@iienet.org or (770) 349-1110.