Mapping service that lets a user explore the car accidents across New Zealand. This will collate other data sets to filter down the results with context.
We originally planned on creating an application that would allow a user to select crash records with contextual filters. We built the application to the point that it was ready for the data, however we encountered a problem when trying to read the data from NZTA. The plan was to use location and time to add context from other sources, like historical rain data, store locations and pubs, and large events happening around New Zealand.
The data from NZTA had been separated into two datasets, one with the date and time, another with location. We could find no accurate way to join the two data sets. Because of this hurdle, we ended up at the end without a database, and only research to show for our efforts.
We considered using a variety of data to create joins between the two NZTA datasets, including suburb (1) as an intersect with precise location (2), light conditions (both), position of accident mid-block/intersction (both), people injured and nature of injuries (both), road conditions (both), and road sealing (both). In the end, the difference in number of records in each database precluded an accurate match being complete.
The sorts of questions that we wanted to examine included:
1) Any relationship between sun-strike position and crash position in space - ie, were some places worse than others through time (sun-strike data by calculation and geographical position)
2) The relationship between rainfall intensity and crash volumes in space and time, testing whether rainfall on its own is the key parameter, or whether intensity of rainfall is more important (rainfall data from HBRC Hilltop server endpoint)
3) The relationship between takeaway shops (Google open source business data) and crashes - testing whether people eating food while driving might be the cause of additional crashes
4) Whether people leaving events and functions (Eventfinda API data and TripAdviser API) were an additional risk for crashes, by mashing event data and crash data
GitHub link - https://github.com/mitchellwarr/Govhack2017-BlueShell
