Human mobility analysis is very useful in urban planning, ecology,
epidemiology and resource planning.
Analysts of this field are accustomed to use call data records (CDR).
Following information can be derived from CDRs:
- Home and work location
- Commute distance
- Calls Per day
- Class Probability and call time
- Hourly locations
For analysis of data there are various techniques available to generate
synthetic CDRs. They are Random Waypoint (RWP) and this technique generates CDR
from random distribution and Weighted RWP (WRWP) which derives CDR by using
weighted distribution of start location and end locations.
Sibren Issacman et al devised a highly accurate method to derive accurate CDR suitable
for analysis yielding highly accurate results as derived from actual CDR or original census data. The name of technique is WHERE means “Work and Home extracted regions” which
starts by computing
distributions for home and work locations, commute distance, calls per
day,
class probability and call time and hourly locations. Then based on
these
distributions WHERE will generate synthetic CDR. This method is highly
accurate as it makes use of more information from actual data instead of
having only a weighted set of inputs for WRWP.
However, as this is based on actual distribution, there is some
probability to derive actual information from the synthetic data. To provide
protection to PII it becomes necessary to use privacy preserving technique like
differential privacy. This is done by a variant of WHERE known as DP-WHERE
(Differentially private WHERE)
In DP-WHERE after getting the distributions we add noise to the
distribution and the use the noisy distribution to generate the synthetic data.
Due to law of the land in most of the countries CDRs are not released
for analysis. If some accurate provable mechanism can be devised and this data
can be brought out then there can be huge potential for research in the field
of urban planning, ecology, epidemiology and resource planning.
Conclusion: A lot of work in
the fields of urban planning, ecology, epidemiology and resource planning is
not being done as the data is not available so there is a huge potential for
data sanitizers to work and provide PII free data to analysts
References
- Sibren Isaacman et al, Human Mobility Modeling at Metropolitan Scales, Proceedings of the 10th International Conference on Mobile Systems, Applications, and Services, http://doi.acm.org/10.1145/2307636.2307659
- Darakhshan Mir et al, DP-WHERE: Differentially private modeling of human mobility, IEEE International Conference on Big Data, http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6691626
