Why cell towers?
Singapore penetration rate is one of the highest in the world (>150%).
Ride on existing infrastructure.
a. No public data on number of cell stations in SG.
b. From local news, 3000 cell stations are required by new teleco TPG to achieve island-wide coverage.
c. From open-source cell information, there could be 50,000 cell stations in SG (across different operators).
d. Accuracy of cell tower to determine position could be vary from 300m.
Using Big Data to Contact Trace.
Cell tower records provide coarse grid location that maps human mobility over time. Associations could be established between unlinked COVID-19 cases where patients who were unknown to each other could have been in the same geographic and time space during their incubation period. This provides meaningful a prior investigation clue to contact tracers so that their interviews with patients could be more targeted. It can also provide early warning to contract tracers should a particular grid location has recorded significant increased visits by COVID-19 patients across time or derive missing links to deduce the presence of asymptotic carriers.
In summary, big data could help
a. Increase Efficiency - Determine mobility of COVID19 patient over past 14 days prior to conducting interview.
b. Link unknowns – Connect known COVID19 cases by across geo-spatial and time slices. Helpful in cases where patients do not know each other.
c. Early detection of potential clusters.
d. Determine presence of asymptotic carriers.
Compliment to existing solutions.
Cell-phone tower records can be integrated to form a more comprehensive and coherent movement picture of individuals to facilitate contact tracing.
- Runs on Android and iOS.
- Feature-phones used by foreign workers or elderly are not supported.
- 10% of population could be running older versions of Android.
- Resistance of citizens to use applications due power consumption issues.
- Difficult to implement for areas without enclosed borders (e.g. hawker center, neighbourhood center, parks).
What needs to be established.
Understand Operational Constraint.
a. Does operators keep base station connection records?
b. Can operators access remotely to retrieve logs from base station daily?
c. Privacy requirements from Personal Data Protection Act? Could we data scramble both user identity and location at respective operator backend to circumvent PDPA?
Establish Technical Challenges.
a. What is the location accuracy at different locations in SG?
b. Will associating records with GIS transport layers, factoring travel speed, improve localization accuracy?
c. What is the data architecture to perform big-data mining supporting geo-spatial analysis? 6 trillion data points for 1-minute location update on 5M cell-phones over 14 days.