Visualizing Crashes and “Risk Areas” for the City of Boston
The Boston Crash Model is a volunteer led project working in partnership with the city of Boston. This project uses data science to understand and predict traffic-related injuries. The following is based on an update from Data for Democracy member Ben Batorsky on the Boston Crash Model Project:
In 2016, the City of Boston as well as other cities around the country and the globe embarked on Vision Zero initiatives with the goal of reducing fatal and serious traffic crashes to zero.
Vision Zero is multi national initiative to reduce fatal traffic accidents. The city of Boston aims to eliminate fatal and serious accidents by the year 2030. In order to start addressing the problem, it is imperative to start tracking accidents. The hope is that data science will make it possible to understand what makes some intersections or sections of the road more dangerous than others. If we can find patterns in the data that shed light on what causes accidents, it’s possible to prevent them.
As part of an effort to support the City of Boston’s Vision Zero initiative, Data for Democracy volunteers created a tool to help the city visualize accident data and identify risky road segments.
The Map
One of the quickest ways to see where the most dangerous intersections are is to visualize the data on a map.
The model predicts, for each road segment and intersection, for each week, whether an accident is likely to happen. Mapping the crash data helps to look beyond pure statistics and get a feel for the human impact of serious and fatal injuries from traffic accidents.
This current map displays crashes that happened in 2016, week by week, alongside predictions from a model developed by the volunteer team. The model itself is based on 2016 crash data and features of the roads themselves. These predictions can be thought of as “risk scores” and will help the city identify areas in need of additional intervention.
Next Steps — Creating More Models
The team will continue improving the models and working with additional data to provide even better estimates of crash risk. We’ll also be be standardizing this method and deploying it in other cities to support traffic safety efforts. Currently we’re looking into applying this to other cities with Vision Zero initiatives and the motivation to tackle the issue of automobile crashes (e.g. Cambridge, MA and Brisbane, Australia).
The code for generating the data, predictions and visualization are all available on Data for Democracy’s GitHub.
Join us on the #p-boston-crash-model channel on the Data for Democracy slack!
