We stumbled across an interesting dataset on data.gov.au: “All vehicular crashes on Tasmanian roads reported by Tasmanian police, or in the case of property damage only, crashes reported by the public.”
The dataset includes approx. 140,000 records, which capture all crashes for a 10 years period, between January 2004 and July 2014.
The first question we wanted to analyse is the number of accidents trend. For this, we generated three graphs to show the evolution of number of accidents per year, organised by severity (it is to be noted that 2014 does not include full year data, and therefore cannot be considered for the analysis).
Note: All chart and map images on this post can be clicked to access interactive chart or map on the opendatasoft platform.
Chart #1: all, property damage only and minor crashes
Chart #2: First aid, Serious accidents
Chart #3: Fatal accidents
Looking at the first chart, the peak in terms of total number of crashes was in 2009, with a steady reduction since then. Although the second chart does not follow a similar pattern, the number of fatal accidents as shown on chart #3 also peaked in 2009, with significant reduction in the following years. We haven’t investigated factors which may have contributed to the reduction of number of accidents, although this could be an interesting topic.
Another questions that we wanted to analyse is whether there are dangerous spots in Tasmania, i.e. areas highly prone to car accidents. For this, we started by generating a heat map of the whole island.
As expected, Hobart and Launceston, given the size of their population and density have the highest concentration of car accidents. We then decided to zoom on Hobart to identify whether specific intersection(s) or road(s) are particularly dangerous. Interestingly, we identified one roundabout which seems to be highly prone to accidents.
Switching to the clustering view, we discovered that 1,500+ accidents occurred on this roundabout, with 938 (!) accidents located at one particular spot, the intersection between Brooker avenue and Liverpool street.
Using “Intersection of Brooker Avenue and Liverpool Street” as a keyword search we then isolated those accidents. The vast majority of them are classified as “Property Damage only”, and the trend is somewhat similar to the general trends of car accidents in Tasmania, i.e. a peak in 2009, and a steady reduction following that year.
Our next analysis was to observe concentration of accidents in Launceston. Although dense, we did not identify one particular “hot spot”.
Finally, we wanted to identify potential hotspots in rural areas. For this we produced a heat map of serious accident, overlaid with a clustered view of fatal accident. We identified, amongst others, a hot spot near Sheffield, country town of 1,400 people, 23km inland from Devonport.
A quick google search “road accident sheffield” confirmed this finding….
With this type of data now being available to the public, we could think of numerous use cases where it could be used in mobile applications and have positive impact on the community. For example, if this data was to be consumed by a road trip planner type mobile app, it could notify users if they were to drive through an accident prone area. Using OpenDataSoft data API, plugging such an application on the standardised restful API would not be too complex….
Published by Alex for Peclet Technology Pty. Ltd. Peclet provides a turnkey open data platform for organisations to share their data. For more information please visit www.peclet.com.au