By Victoria Woollaston Daily Mail
Where do you see yourself in five years time? It’s a common
interview question designed to learn about a person’s ambitions, and thanks to
new technology you could soon be able to give a precise answer to this question.
Researchers from Microsoft and Google can predict where a
person will be years from now using a new computer software called Far Out.The
programme tracks a person using a GPS device and learns their routine.
It then uses this information to accurately guess their
future locations and will adapts its predictions even if someone changes their
job, relationship or moves house.
HOW DOES FAR OUT WORK?
Using GPS systems carried by volunteers and fitted to the
transport they used on a daily basis, the researchers were able to plot around
150 million location points.
Furthermore, over 32,000 days worth of precise GPS data was
collected.
This information was fed into the Far Out software which
uses an algorithm to predict where a person will be in the future, based on
where they’ve been in the past.
It does this by accurately ‘learning’ a person’s routine.
Far Out then offers predictions but can also automatically
discover when someone veers from this routine.
The program will plot these changes, learn from them, and
adapt accordingly.
Researchers believe the results could be used to predict
rises in populations, the spread of disease, traffic and broadband demand to
quote a number of the researchers examples.
Adam Sadilek and John Krumm wanted to learn more about what
they call ‘human mobility’ so set about developing a formula that could predict
where a person would be years in the future in their paper ‘Far Out: Predicting
Long-Term Human Mobility.’
They gave 703 volunteers in Seattle a GPS device and told
them to carry it around with them at all times, including going to work,
shopping, out with friends and even travelling.
The researchers also fitted GPS units to the buses, cars and
other transport used by the volunteers on a daily basis.
Sadilek and Krumm managed to collect around 150 million
location points and 32,000 days worth of GPS data during this experiment.
This data was fed into their Far Out computer program to
predict the ‘long-term human mobilty’ of each subject.
Far Out was able to offer predictions, and could also
automatically discover when someone veered from their usual routine, adapting
accordingly.
‘For example, [Far Out] might notice that Tuesdays and
Thursdays are usually about the same and fairly consistent from week to week,’
the researchers told Fast Company.
‘Then when we ask about a future Tuesday or Thursday, the
algorithm automatically produces a typical Tuesday/Thursday as a prediction.’
‘If there is a sharp transition, such as a move to another
city, the system notices there is a discrepancy between its predictions and
actual data and adapts to the new patterns.
’Most people have
only a few ‘revolutionary’ changes in the course of their lives, so Far Out
isn’t caught off guard too often.’
According to Salidek and Krumm’s paper, humans are mostly
predictable even over extended periods of time.
The researchers gave 703 volunteers in Seattle a GPS device
and told them to carry it around with them at all times.
The researchers gave volunteers in Seattle a GPS device and
told them to carry it around with them at all times. The researchers also fitted
GPS units to the transport used daily by the volunteers. Sadilek and Krumm
managed to collect around 150 million location points and 32,000 days worth of
GPS data
The GPS data was fed into the Far Out system so the program
could learn a person's routine.
The GPS data was fed into the Far Out system so the program
could learn a person’s routine. Far Out then predicted where that person would
be, including their longitudinal and latitudinal positions, on a certain date
in the future. The software even takes into account if there is a national
holiday on that date
The paper said: ‘Much work has been done on predicting where
is one going to be in the immediate future, typically within the next hour.
By contrast, we address the open problem of predicting human
mobility far into the future, a scale of months and years.’
Far Out uses ‘a nonparametric method that extracts
significant and robust patterns in location data, learns their associations
with contextual features (such as day of week), and subsequently leverages this
information to predict the most likely location at any given time in the
future.’
The researchers believe the software could have social uses,
similar to Foursquare, or could be used by advertisers and marketers.
They also believe the results ‘open a number of interesting
avenues for future research and applications.’
This could include predicting rises in populations, the
spread of disease, traffic problems and broadband demand.
No comments:
Post a Comment