Modeling Geographic Patterns With Mobile User Location Data

One of the things we (meaning the LBS community) talked about in the early days of 911, was using call log histories to improve wireless coverage.  The idea is simple.  Take identity-ridden, non-intrusive logged caller locations (which include signal strengths) and create an interpolated GIS model to then isolate under served areas denoted by weak signals in a continuous field- red being strong, yellow being weak.  We saw this GIS modeling work as an opportunity to replace expensive drive-by systems used at the time by mobile operators to improve their coverage and optimize service.  While the idea was noble, it never took.  All the carriers we worked with just let the data hit floor and it was swept away into dust bins or saved only for odd court-issued subpoenas.  What a waste.

Since then, lots of folks have caught on to the idea that anonymous logged mobile locations in any transaction context can be used for all sorts of modeling and new data creation.  From traffic models to isolating target-rich advertising zones, modeling based on post-transaction analytics and business intelligence is the new trick of the Google era. 

Researchers at Northeastern University are with it (I think most around the world are.  City University in London was doing this in the late 90s).  Northeastern recently used 100,000 anonymous mobile locations to map social patterns of geographic interaction.  No surprise - we humans are for the most part sedentary creatures, staying within 20 miles of our homes (begs the question why we need TomTom's!).  Researchers hope to extend the data findings into epidemiological analyses and use it in a similar context to John Snow's famous cholera outbreak analysis map of central London produced in 1854.  It's good to see this advanced GIS work finally happening with LBS...