THP Aims to Take the Guesswork Out of Traffic Safety With Predictive Analytics
Tennessee is known as the Volunteer State, so it comes as no surprise that the Tennessee Highway Patrol (THP) has become the first highway patrol and state police agency to voluntarily apply IBM's predictive analytics model to traffic safety. THP analysts plug an array of factors into the software, and the program then gives predictions of when and where serious traffic accidents are most likely to occur. Troopers can then be deployed to problem spots ahead of time, either to deter the predicted crashes from occurring in the first place or to be on the scene of an accident more quickly to render better assistance.
Spearheading the program, officially dubbed Crash Reduction Analyzing Statistical History (or C.R.A.S.H.), has been Colonel Tracy Trott. This 37-year veteran of the THP was promoted to colonel in the fall of 2010. More recently, he was elected as the 2nd Vice Chair of the State and Provincial Section of the International Association of Chiefs of Police (IACP). He sat down with us recently to discuss predictive analytics technology and how successful the THP has been in implementing it.
IACP: What prompted the Tennessee Highway Patrol's (THP's) decision to use data analytics in its efforts to reduce traffic accidents and fatalities?
COLONEL TRACY TROTT: Two things. One, our assistant commissioner was a man named Larry Godwin, who was the former director of police services for the Memphis (Tennessee) Police Department. They had used predictive analytics on the issues of violent crime and property crime in Memphis, and they had been pretty successful at implementing that program and lowering their crime rates. He had an association with IBM.
IACP: And the second thing?
TROTT: The real motivating factor was this. Nobody is going to walk up to me tomorrow and give me 200 new troopers. Budgets are tight, and new money is in short supply. So, my job is to figure out how to make us better and how to work smarter with what we've got. We started to look at the predictive analytics model. We met with IBM, and we asked them if they thought that this could be applied to traffic safety and reducing traffic fatalities. No one had ever done that before. In fact, we're the first state police department in the United States that has taken a predictive analytics model and applied it to traffic safety. They said it could be done depending on what data we had available to us to plug into the system. We entered into a contractual agreement with IBM and started our program in January 2014. For the first six months of last year, we did a pilot project in three different districts. Then, in June of 2014, we went live in all eight of our districts.
IACP: What kind of data has been plugged into the software in order for the technology to make its predictions as to when and where serious accidents are likely to occur?
TROTT: We not only take crash history that we have in our system through electronic reporting, we also have plugged in weather factors two weeks out from the National Weather Service. We'll also take special events that our captains in the districts know about and plug them in. Like if they're having a Mule Day Festival in Columbia, Tennessee, that's going to attract a lot of people or if the University of Tennessee is playing a home football game in Knoxville and there's going to be 100,000 fans in town. We also key in any arrest history as far as DUI-related crash history.
IACP: Do you have any statistics to share with our readers as to how accurate the program has been?
TROTT: After the first six months of model programs and then the six months of 2014 that we went live in all districts, we went back and overlaid the actual incidences and compared that to the predictions the model made, and the model was accurate 70 to 72 percent of the time. So, we feel our model is pretty well designed if we're getting that kind of prediction accuracy.
IACP: Some say this new tool can be as revolutionary as seat belts and radar. What do you say?
TROTT: Well, I don't want to throw all of our eggs into one basket. We want to be multi-faceted. We want to be very good at our DUI enforcement. We know that is a major cause of crashes. We want to be very good at our seatbelt enforcement and get our compliance rate for seatbelts way up. But we want to use this new tool to augment all of that and really direct our resources. Traditionally, highway patrol and state police agencies have been very reactive. There is a wreck, you go the scene, you take care of the scene. Maybe you see a pattern of problems and you flood people in those areas at certain times, and you try and solve things that way. We really want to get to where we are proactive and try and do something about these tragedies before they happen. And if we can't do that, at least we can try and be in the area where they are most likely to happen for two reasons. One, maybe our enforcement action or presence can prevent some of them from happening. And, two, we'll be able to take care of emergencies much more quickly, much more efficiently, and maybe lessen the impact of each.
IACP: What are the advantages of working with this particular tool?
TROTT: IBM has been a great partner for us. We employ one of their business solutions architects to help us build the dashboards that are going to be available to all of the troopers. That's where the success of the program is really going to come from. When we get to the point where our troopers go to work one day, they pull up the predictive analytics model, and it shows a 65 percent probability rate of a crash being in a certain area of their county at a certain time of day, and they self-deploy themselves to that area. We're just now in the infant stages of the program. We're trying to get people to believe in it and buy into it. Most of our upper command has done that. We're now designing our enforcement programs based on the predictive analytics models. And as people get more experienced and see the success of the model, then it will push itself down lower into the ranks where it will really be a success. This past year in Tennessee was the second lowest traffic fatality year since 1963. That was a full year of predictive analytics. It speaks for itself.
IACP: And with this being such a pioneering use of predictive analytics, how much convincing did it take at the various levels of bureaucracy and the chain of command to give this technology a chance?
TROTT: On the upper part of the organization, it was an easy sell. [Commissioner of the Tennessee Department of Safety and Homeland Security] Bill Gibbons is a very data-driven man. I think Governor Bill Haslam relies on data to direct his activities and his priorities. So, it took very little convincing to apply this to the Tennessee Highway Patrol. My job was to sell it to the uniformed people of the highway patrol, who traditionally had operated on their own knowledge and their own feelings about where they needed to be and what they needed to be doing. We did this through a number of educational opportunities. We went out into the field to train captains, lieutenants, sergeants, and troopers on the use of the program and show them the benefits of deploying themselves in certain areas with high probabilities and in certain times and show them all of the inputs going into the predictive model. I think we've achieved what we wanted to achieve. We wanted to get our district commanders and district supervisors to buy into the program. I think we've done that very well. I think right now we're only scratching the surface of what we can do with it and how effective it will be.
IACP: What was the cost factor?
TROTT: We were able to make the initial investment with IBM for the predictive model through a federal grant that we got through the Governor's Highway Safety Office (GHSO). Kendell Poole is the Director of the GHSO, and they help us in so many ways. Their mission, like ours, is to save as many lives as they can. They graciously helped us fund this project with the initial investment. They also helped us in the employment of analysis. We had to have people that took the data spit out by the program and put it in a form where our captains can retrieve it and understand it.
IACP: If we were to chat a year from now, what would you hope to report as a result of this technology being implemented?
TROTT: We decreased fatalities around 3 percent this past year. The last 4 years have been 4 of the 5 lowest fatality years in 50 years. We're on a good path, and we have a program in place that is showing success and leads me to believe we'll have future success. We've also had a lot of interest from other states that have heard about our program. I think, in years to come, we'll look back and say this was probably a watershed moment for the Tennessee Highway Patrol and maybe all state police agencies across the United States.
Abstract News © Copyright 2015 INFORMATION, INC.