By Robbie Webber
The Tennessee Highway Patrol has begun using software that will predict where crashes and other safety problems will occur. However, instead of simply identifying problem locations over the long term, the model looks at four-hour segments and 30 square mile areas. This allows police officers and resources to be efficiently dispatched to specific areas to either prevent or respond to anticipated high-risk situations.
The Crash Reduction Analyzing Statistical History, or CRASH, program uses multiple data inputs, including special events—sporting events, festivals, holidays, etc.—weather, and historical data. In the six months it has been running, CRASH has had an accuracy rate of approximately 72 percent. The THP is still assessing its performance, but traffic fatalities are down about 5.5 percent from this time last year. Although veteran police officers have been predicting trouble spots for years, the program allows many more factors to be considered and puts science behind the predictions.
CRASH grew out of a similar modeling of crime in Memphis—Blue CRUSH—that helped target hot spots for forcible rape, gun violence, and other crimes by analyzing data and finding the patterns more effectively than humans were able to.
In addition to the crash-focused model, the THP is also deploying a model aimed at predicting where and when drivers who are under the influence of drugs or alcohol will be on the road. One of the factors that program considers is the location of places that sell alcohol under an Alcoholic Beverage Commission license.
Utah DOT is also using a predictive model developed at Brigham Young University to allocate funding to specific locations deemed to be high risk, however the Utah program focuses on infrastructure changes rather than on circumstances that might change from day to day.
Robbie Webber is a Senior Associate at SSTI.
By Robbie Webber