Correlations of Traffic Speeds with Crash Incidents / Chicago, IL

Project Name: Correlations of Traffic Speeds with Crash Incidents / Chicago, IL

LOADING PROJECT

How to Operate:
To operate this visualization, click play to use the default run configuration.

 

From here you can adjust the playback speed, zoom to different points in the city, and hover over streets and accidents to see their individual data and attached metadata.

 

There is also the option to change the cumulative duration of the points on the map using the slider at the bottom of the window. You may adjust this window to display all points within the specified time frame.

 

Project Description:

This is a visualization correlating real time traffic speed data for segments of Chicago arterial streets with real time accident reporting, with the purpose of correlating fluctuations in traffic speed to an increase in accidents.

 

A little explanation about the data itself, the traffic data is generated by the Chicago Traffic Tracker (CTT), where this system estimates traffic congestion for around 1,250 segments covering 300 miles of of Chicago’s arterial streets in real-time by continuously monitoring and analyzing GPS traces received from Chicago Transit Authority (CTA) buses.

 

The traffic accident data was taken from the Chicago Police Department (CPD), where the data is recieved as is from the electronic crash reporting system (E-Crash) at the CPD, excluding any personally identifiable information.

 

The goal of this project was to show that fluctuations in the speed of a road segment could be correlated to an increase in accidents. I started by generating a visualization of road speeds, and then added the reported crashes within approximately 22 meters of streets that contained valid data.

 

After examining the map in more detail, it became apparent that although crashes occurred all over the city, within this 4 day period they were mainly centered on major streets spanning Chicago from north to south.

 

Additional Items:

After graphing the crashes per arterial street, this was also able to be confirmed for this time period.

 

 

Through this information, I was able to choose a street to profile in depth, Halsted, as the accident count was high and this is a street I have had to drive myself during peak congestion hours. Additionally, it became apparent from the City wide map that this street would have a decent spread of crash data along different segments.

 

I then realized that there were disparities between the dataset’s schemas, and saw that some data processing would be needed to accurately relate the elements I desired. In this case, metadata was and still is available from a number of sources to aid in this endeavor, namely from Terbine.io and the City of Chicago’s Data Portal. As a result, the data processing was made much easier, as I was easily able to accurately find data types and meanings of data fields, which is not always the case with sensor generated data.

 

After plotting the speed for each street segment, I needed a way to visualize the variation between the data points. I settled on using the Mean absolute deviation (MAD), to reduce the dispersion caused by extreme outliers.

 

Combining this information with the crashes located on each segment individually, we get the following graphs.

 

 

From this, it can be seen that, on average, when an accident occurred there were large fluctuations in the road speed at the time.

 

Real time data processing can be used to make these correlations as data is available, allowing the City to better address road construction, policing, traffic light timings, and assessment high risk areas on small and large scales dynamically. With this information, the government could react in real time to the needs of the people commuting and living in these areas to better address their safety through infrastructure and smart development where it is needed most.

Developer(s): Nicholas VanCise, University of Nevada-Las Vegas

 

Data Feeds Employed:

 

Open Source Tools:

kepler.gl

Matplotlib

 

GitHub: https://github.com/Terbine/projects/tree/master/ctt_crash_visualization

 

Original Posting Date: 9 September 2020