D4D Project Highlight: The Data for Democracy Crash Model

Last year we published an update on the Boston Crash Model, a volunteer led project in partnership with the city of Boston. Since then, the focus of the project has expanded to more cities and seeks to continue growing beyond Boston.

Introducing Insight Lane: the Data for Democracy Crash Model

by Jenny Turner-Trauring and the Insight Lane team

Data for Democracy is proud to introduce the latest release of Insight Lane (formerly Boston Crash Modeling), a data analytics project designed to help cities understand and manage risk on their roads.

Unlike the initial release which was specific to Boston, this release offers a pipeline process that can standardize any city’s raw data, generate predictions and display them via a web visualization tool.

Project Goals

As a data analytics project, we aim to provide metrics that support the decision-making of cities’ transportation departments as they carry out their mission to reduce traffic crashes. The specific decisions that can benefit from these metrics are:

  1. Identifying High Risk Locations — which locations in the road network represent the highest risk of crashes?
  2. Understanding the Contributing Factors of Risk — what features, patterns and trends can be identified that contribute to a location’s risk of crashes?
  3. Assessing the Impact of Intervention — what is the impact of a proposed (or past) intervention on the risk of crashes?

From Boston to multiple cities

Photo by STANLEY NGUMA from Pexels

Partnering with the City of Boston in 2017, our volunteers used publicly available maps and crash data to build a tool to visualize and predict the likelihood of crashes.

Boston is just one city, however, and we believe that any city working to reduce its crashes can benefit from what our tool has to offer. Many cities are adopting open data ordinances, and publishing crash data and other resources. With our latest release (and new name!) we are expanding the scope of our project to support multiple cities.

We started by adding two new pilot cities to our Boston demo: Cambridge, Massachusetts, and Washington, DC. Insight Lane can be used by any city with crash data, but we are using these cities to demonstrate our tool’s range. Here’s how the project has changed to accommodate new cities:

  • OpenStreetMap: We now use OpenStreetMap, a free, open, crowd-sourced map, to build the road network, instead of city-supplied maps. This gives us a consistent starting point across cities that includes many great predictive features, such as road width, number of lanes, traffic signals, crosswalks, bus routes, bike and pedestrian facilities.
  • Resident concern data: Our pilot cities — Boston, Cambridge, and DC — all have some form of concerns reported by residents. For Boston and DC, these concerns can be submitted via their Vision Zero portal, and focus on traffic safety concerns such as people speeding or a road with too many lanes for pedestrians to cross safely. Cambridge does not yet have a Vision Zero portal, but has See Click Fix, a portal where people can submit other categories of concerns; some are related to traffic, but most are not. Although the Vision Zero safety information is more highly correlated with crashes, even locations with a large number of See Click Fix concerns correlate with a higher number of crashes. We include the number of concerns at a location as one of our features.

Going forward, there are additional data sources we hope to add, with the goal of supporting multiple cities:

  • More resident concern data: Many cities have 311 data which can be used as their concern data.
  • Traffic volume: Traffic volume is a very useful feature when predicting crash risk. Because Boston was our first pilot city, we incorporated their traffic volume in the form of Automated Traffic Recording counts and Turning Movement Counts. Unfortunately, cities have a wide range of formats in which they make their traffic counts available. We hope to work with cities to come up with a standard format for traffic counts, but until we do, adding traffic volume will have to remain a custom solution built for each city.
  • More features from OpenStreetMap: We are not yet taking advantage of all the possible features that OpenStreetMap makes available. We plan to add as many as we can.
  • Other data sources: There are also a number of other publicly available data sources that we plan to utilize, including traffic citations, transit schedules, and weather.

Future work

Photo by Isaque Pereira from Pexels

Beyond adding additional data sources, our short term goals are making Insight Lane more informative and user-friendly:

  • Improved reporting: Provide lists of highest risk road segments, as well as road features that most closely correlate with risk.
  • Before & after assessments: Assess the impact of specific interventions undertaken by traffic departments. Some examples of interventions are bike lanes, raised crosswalks and new traffic signals.
  • Better UX: Develop a UI tool to more easily enable cities to run our pipeline to visualize and analyze their data.

Get involved!

  • You can visualize the data for our pilot cities here
  • You can also use our pipeline to take your city’s publicly available crash data and to visualize that city’s crashes and risk data. To learn more about how to run the pipeline, check out our README on GitHub.

If you are interested in this project or others like it, join us and get involved in the #p-boston-crash-model channel.