Visualizing accidents in France from 2005 to 2020 , originally released as French Government open data .
The project is developed as part of the Information Retrieval and Visualization lecture lecture at Martin Luther University Halle-Wittenberg.
Start a web server with the visualizations running on http://localhost:8080/:
yarn start
Note that you must first preprocess the data.
If you want to deploy the compiled, static resources to a HTTP server, run
yarn build
and copy the dist/
folder to your web server's content root.
In order to fix various encoding and label issues and to combine the different open datasets, we need to preprocess the data once before starting the web server. Here are the steps required:
- Install Python 3
- Install pipx
- Install Pipenv
- Install dependencies (this may take a while):
pipenv install
- Run preprocessing:
pipenv run python preprocessing/preprocess.py
To randomly sample a smaller test dataset for testing purposes, run the following:
shuf -n 10000 static/data/accidents.jsonl > static/data/accidents-sample.jsonl
- time series
- x-axis: time
- interaction: strip year or not (might be possible to detect yearly trends)
- y-axis: number of injuries/fatalities
- interaction: switch between injuries and fatalities
- banking to 45 degrees
- interaction: toggle color scale of line:
- none
- birth year
- interaction: filter for different safety equipment
- icon-based (stick figures) für personen
- filter for drivers/passengers
- maybe more filtering
- young vs. old drivers
- sex
- reason for travelling
- safety equipment
- alone vs. accompanied by
- plot in geographical coordinates
- grid every ?? kilometers
- interaction: change aggregation type
- average values and plot one single stick figure per grid cell
- plot each incident as one stick figure but overlay with x-ray technique
- that way no detail is lost by aggregation
- interaction: additional color dimension
- proportion of persons injured or dead
- treemap (or tree) of accident types:
- collision types
- road category
- light conditions
- weather
- intersection type
- road curvature
- vehicle type
- situation
- interact: switch between tree and treemap
- Are there repeating yearly patterns? Are roads more dangerous in winter?
- Do older or younger drivers drive more safely?
- Where are more / the more severe accidents? city or rural areas?
- Do the proportion of dead and injured persons correlate? Where are they most different?
- Do dedicated bicycle lanes make roads safer for cyclists?
This project is MIT licensed, you can use it for whatever you want as long as you mention this repository. We use the Elm framework which is also open source.