1. Measuring and Predicting
Departures from Routine
in Human Mobility
Dirk Gorissen | @elazungu
PyData London - 23 February 2014

2. Me?
www.rse.ac.uk

3. Human Mobility - Credits

University of Southampton





BAE Systems ATC


James McInerney
Sebastian Stein
Alex Rogers
Nick Jennings
Dave Nicholson
Reference:


J. McInerney, S. Stein, A. Rogers, and N. R. Jennings (2013).
Breaking the habit: measuring and predicting departures from
routine in individual human mobility. Pervasive and Mobile
Computing, 9, (6), 808-822.
Submitted KDD paper

4. 
Beijing Taxi rides

Nicholas Jing Yuan (Microsoft Research)

5. Human Mobility

London in Motion - Jay Gordon (MIT)

6. Human Mobility: Inference

Functional Regions of a city

Nicholas Jing Yuan (Microsoft Research)

7. Human Mobility: Inference

Jay Gordon (MIT)

8. Human Mobility: Inference

Cross cuts many fields: sociology, physics, network
theory, computer science, epidemiology, …
© MIT
© PNAS

9. Project InMind

Project InMind announced on 12 Feb

$10m Yahoo-CMU collaboration on predicting human needs and
intentions

10. Human Mobility

Human mobility is highly predictable



Average predictability in the next hour is 93% [Song 2010]
Distance little or no impact
High degree of spatial and temporal regularity



Spatial: centered around a small number of base locations
Temporal: e.g., workweek / weekend
“…we find a 93% potential predictability in user mobility
across the whole user base. Despite the significant
differences in the travel patterns, we find a remarkable
lack of variability in predictability, which is largely
independent of the distance users cover on a regular
basis.”

11. Temporal Regularity

[Herder 2012] [Song 2010]

12. Spatial Regularity

[Herder 2012] [Song 2010]

13. Breaking the Habit

However, regular patterns not the full story



travelling to another city on a weekend break or while on
sick leave
Breaks in regular patterns signal potentially
interesting events
Being in an unfamiliar place at an unfamiliar time
requires extra context aware assistance


E.g., higher demand for map & recommendation apps,
mobile advertising more relevant, …
Predict future departures from routine?

14. Applications




Optimize public transport
Insight into social behaviour
Spread of disease
(Predictive) Recommender systems





Based on user habits (e.g., Google Now, Sherpa)
Context aware advertising
Crime investigation
Urban planning
…
Obvious privacy & de-anonymization concerns
-> Eric Drass’ talk

15. Human Mobility: Inference

London riots “commute”

16. Modeling Mobility

Entropy measures typically used to determine regularity in
fixed time slots



Well understood measures, wide applicability
Break down when considering prediction or higher level structure
Model based






Can consider different types of structure in mobility (i.e., sequential
and temporal)
Can deal with heterogeneous data sources
Allows incorporation of domain knowledge (e.g., calendar
information)
Can build extensions that deal with trust
Allows for prediction
Bayesian approach


distribution over locations
enables use as a generative model

17. Bayes Theorem

18. Bayesian Networks


Bottom up: Grass is wet, what is the most likely cause?
Top down: Its cloudy, what is the probability the grass is wet?

19. Hidden Markov Model


Simple Dynamic Bayesian Network
Shaded nodes are observed

20. Probabilistic Models


Model can be run forwards or backwards
Forwards (generation): parameters -> data

E.g., use a distribution
over word pair
frequencies to
generate sentences

21. Probabilistic Models


Model can be run backwards
Backwards (Inference): data -> parameters

22. Building the model




We want to model departures from routine
Assume assignment of a person to a hidden location
at all time steps (even when not observed)
Discrete latent locations
Correspond to “points of interest”

e.g., home, work, gym, train station, friend's house

23. Latent Locations


Augment with temporal structure
Temporal and periodic assumption to behaviour


e.g., tend to be home each night at 1am
e.g., often in shopping district on Sat afternoon

24. Add Sequential Structure

Added first-order Markov dynamics


e.g., usually go home after work
can extend to more complex sequential structures

25. Add Departure from Routine


zn = 0 : routine
zn = 1 : departure from routine

26. Sensors

Noisy sensors, e.g., cell tower observations


observed: latitude/longitude
inferred: variance (of locations)

27. Reported Variance

E.g., GPS

observed: latitude/longitude, variance

28. Trustworthiness

E.g., Eyewitness


observed: latitude/longitude, reported variance
inferred: trustworthiness of observation

single latent trust value(per time step & source)

29. Full Model

30. Inference

31. Inference is Challenging


Exact inference intractable
Can perform approximate inference using:

Expectation maximisation algorithm



Gibbs sampling, or other Markov chain Monte Carlo



Fast
But point estimates of parameters
Full distributions (converges to exact)
But slow
Variational approximation


Full distributions based on induced factorisation of model
And fast

32. Variational Approximation

Advantages





Straightforward parallelisation by user
Months of mobility data ~ hours
Updating previous day's parameters ~ minutes
Variational approximation amenable to fully online
inference
M. Hoffman, D. Blei, C. Wang, and J. Paisley.
Stochastic variational inference. arXiv:1206.7051,
2012

33. Model enables

Inference





Exploration/summarisation


location
departures from routine
noise characteristics of observations
trust characteristics of sensors
parameters have intuitive interpretations
Prediction


Future mobility (given time context)
Future departures from routine

34. Performance

Nokia Dataset (GPS only) [McInerney 2012]

35. Performance

36. Performance


Synthetic dataset with heterogeneous, untrustworthy
observations.
Parameters of generating model learned from OpenPaths
dataset

37. Performance

38. Implementation

Backend inference and data processing code all python




UI to explore model predictions & sanity check





numpy
scipy
matplotlib
flask
d3.js
leaflet.js
kockout.js
Future


Gensim, pymc, bayespy, …
Probabilistic programming

39. Map View: Observed

40. Map View: Inferred

41. Departures from Routine: Temporal

42. Departures from Routine: Spatial

43. Departures from Routine: Combined

44. Departures from Routine

45. Conclusion & Future Work

Summary
 Novel model for learning and predicting departures from routine

Limitations


Need better ground truth for validation
Finding ways to make the model explain why each departure
from routine happened.


Needs more data (e.g., from people who know each other, using
weather data, app usage data, …).
Future Work

Incorporating more advanced sequential structure into the model





e.g., hidden semi-Markov model, sequence memoizer
Supervised learning of what “interesting" mobility looks like
More data sources
Online inference
Taxi drivers

46. Questions?

Thank you.


dirk.gorissen@baesystems.com
| @elazungu
Reference:

J. McInerney, S. Stein, A. Rogers, and N. R. Jennings (2013).
Breaking the habit: measuring and predicting departures from routine
in individual human mobility. Pervasive and Mobile Computing, 9, (6),
808-822.