This New Scientist article reminds me of a project I worked on about ten years ago. The article talks about analyzing mobile-phone location data to establish patterns of how the users move around. We didn’t analyze predictability rates, but we did look for patterns that foretold other patterns, like this bit from the article:
“Say your routine movement is from home to the coffee shop to work: if you are at home and then go to the coffee shop it’s easy for me to predict that you are going to work,” says co-author Nicholas Blumm.
We similarly looked for patterns that predicted that you were on your way home, to work, or some other such. We used that to set up some experimental stuff, like adjusting my house thermostat (which I can control over the Internet) when I was on my way home. It was fun stuff to play with.
At the time, I wrote a program for the BlackBerry that would, if you installed it and enabled it, send telemetry to a “context server” that we created. The telemetry included information about the cell to which the BlackBerry was connected, and it was sent at intervals, and also whenever the cell changed. And we had a mapping of the cell IDs to the actual locations of the cells. We would never get that today, unfortunately, but we could learn the mapping for the user’s local area. And by aggregating the information from many users, we could probably made a pretty good mapping ourselves.
One of my colleagues visited family in northern Virginia, and had his BlackBerry sending telemetry on the way home, on a Monday. I remember mapping that in real time, and watching his progress up US 15, I-78, and I-287. I could even tell where he’d made significant stops, and we had a good time going over the trip when he was in the office on Tuesday.
That was a fun project, and we got some interesting results, though we ultimately didn’t go very far with them. It’ll be nice to see where this new group of researchers goes.
Maybe we’ll track them on their mobile phones.