Commissioner David W. Purkey of the Tennessee Department of Safety and Homeland Security and Colonel Tracy Trott of the Tennessee Highway Patrol (THP) are proud to announce the release of the department’s lifesaving, predictive analytics software to all 95 county sheriffs’ offices in Tennessee free of charge. The predictive models have proven to be 70% accurate where alcohol, drug, and crash incidents may occur.
The Tennessee Department of Safety and Homeland Security’s predictive analytics program has delivered an improved suite of tools for those making resource deployment decisions.
The goal of the program is to reduce traffic crashes and improve crash outcomes. Historical data is used to predict how likely a particular kind of incident could occur in an area at a given time.
Each week, current weather forecasts and upcoming events are passed through the model and the crash forecasts for the following week are generated.
Forecasts are formed based on the date, time, and location of historical traffic crashes in combination with current weather forecasts and knowledge of upcoming events that are likely to affect traffic volume or traffic safety in a significant way. From there, data is sent to the department’s Geographic Information System (GIS) Office where it is converted into interactive maps, one for each day of the upcoming week. In addition to the crash forecasts, the map contains information regarding past crashes, dates and locations of events likely to have an impact on traffic safety.
Sheriffs’ offices can now use both forecast and the historical data to guide them to the places where they are likely to have the greatest impact on traffic safety.
The tool can be used to determine when and where to conduct grant-funded activities, where law enforcement should be during unobligated patrol time, and to assist supervisors when developing enforcement plans for the upcoming week.
A trooper or deputy sheriff can examine the area of their assignment, and conduct patrol and enforcement activities in the areas at the times when the model suggests the risk for a serious crash is highest.
Likewise, supervisors can schedule their available resources other enforcement activities during times and in areas indicated by the model for the upcoming week.
In the fall of 2013, the Tennessee Highway Patrol began work on its first predictive model. Using techniques developed in the field of crime modeling and forecasting, the first “crash” model
predicts the likelihood of fatal and serious injury crashes. Using this information, troopers determined when and where to deploy based on the forecasted risk of serious crashes in their area.
The theory being that their presence can prevent a crash from occurring, or produce a better crash outcome by reducing the response time in the event that a serious or fatal crash does occur.
As traffic fatalities nationwide continue to increase, the fatality rate in Tennessee has decreased from 1.47 per 100 million vehicle miles traveled in 2010 to 1.35 in 2016.
Furthermore, preliminary data indicates that 2015 had the second fewest traffic fatalities in Tennessee since 1963. When crashes occur, THP has reduced its average response time by 33% since 2012. As the department continues to push the predictive analytics program forward, we hope to find creative new uses for predictive modeling of traffic safety data, and continue to improve the safety and the lives of Tennessee citizens and the motoring public.