Data science in govtech
Show different areas of daily work, size
Bridge between IT and policy, data driven decision making.
Required diverse teams
- software engineers
- machine learning
- data needed for public policy
- statistics
- visualization
Understanding library borrower profiles
- from 20m NLB records
- clustering to identify profiles
- found out there are two subgroups of elderly borrowers
Hansard: unsupervised machine learning to uncover topic clusters
- applied to parlimentary speeches
- topics covered, how they are related
Topic modelling to understand public feedback
- 100k emails on HDB flat sales
- cluster of emails about key collection
- HDB implemented a website to allow buyers to change the date to collect their keys
Solving the Circle line mystery
- 250 data points