A new app reveals the locations of frequent serious traffic accidents on Czech roads. The software, developed by scientists from the Faculty of Economics and Administration of MU in cooperation with Charles University will help traffic police officers to more easily implement preventive measures. Michal Kvasnička, its main designer, talks about why the application is needed and how it is unique.
Around five hundred people die on Czech roads every year and many more are injured. The task of police officers is not only to address traffic accidents but also to prevent them. To do this, they need to find dangerous places on the road network. It might seem that all they have to do is look at a map and find the places where traffic accidents are concentrated. But it is not that simple. There are too many accidents on the roads to analyse with the naked eye. We are helping the Traffic Police Service to solve this problem. Our app searches for clusters of serious accidents that signal a possible danger spot. We are therefore helping the Police Service to make use of the data it obtains during standard traffic accident investigations.
Our solution has several interesting features. The most important is that we can analyse the entire road network, including intersections, which are frequent locations of traffic accidents. We can assign different weights to individual accidents according to their severity. If all accidents were given the same weighting, then parking lots would look extremely dangerous because there are a lot of accidents there, however, they do not usually have serious consequences.
Police routinely analyse accident data and use tools to visualise them on maps in the form of a clearly defined point on the road. But we assume that the location of a traffic accident is always at least a little random. For example, the accident could just as easily have happened ten metres further to the left or right. However, the probability that it would have occurred a hundred metres from the actual accident site is already much lower.
Every traffic accident that has ever happened in the Czech Republic sits at the top of an imaginary probability hill. As we move away from the actual accident site on the road, or as we go downhill, the probability that an accident could happen there too gradually decreases to zero. We calculate such a probability distribution for all traffic accidents. We then add up the probabilities, or more precisely the kernel densities, for each road segment across the accidents. Where the resulting sum of densities is high, we identify a cluster of accidents – i.e., a potentially dangerous location in the road network.
The easiest way is to try it out. We’ve published a simplified version of our app for the public on our website at https://trafficacc.econ.muni.cz/, where anyone can look at their favourite road or junction. The app contains data for the whole Czech Republic. First you need to select a district. On the tab called “Clusters of accidents”, the user can then browse potentially risky places. The calculation of accident clusters has many parameters, and there is no single correct setting that would suit, for example, cities, traffic on expressways or third-class roads equally well. To save data, the sample version of the application contains hazardous locations calculated with only one setting, which significantly accounts for fatal accidents and serious injuries. Police officers have a full version of the app in which they can change these settings to get a multidimensional overview of dangerous locations across the country.
The method used proved to be extremely computationally intensive. So we ended up splitting the software into two parts. The first part prepares the data, so it doesn’t matter too much that it runs for several days even on a more powerful computer. This pre-prepared data is then displayed by the web application, which as a bonus also generates basic statistical reports for the districts and allows interactive work with data on individual traffic accidents.
We expect that they will use the app in several ways. Managers can use it to quickly track the number and causes of accidents in their district. Police analysts can then either look at individual accidents at locations of interest to them or at hazardous locations highlighted by the app. The expert knowledge of police officers is crucial when evaluating information from the app. Firstly, it allows them to find the appropriate settings for detecting clusters of accidents, but most importantly it allows them to interpret the results correctly.
Imagine that the app finds a cluster of serious traffic accidents at a certain location. The police analyst can look up the fact that the accidents happened because of a road reconstruction that they know is already completed. These accidents can be ignored. But if they find a risky cluster of accidents that does not have such a clear and completed cause, then they can analyse the individual accidents and discover the cause. The police analysts can then suggest measures to reduce the risk at that location. Our application can therefore be a useful tool, but the role of the expert knowledge of police analysts is still irreplaceable.
The application was developed for the needs of the Police of the Czech Republic through the implementation of a project funded with state support from the Technology Agency of the Czech Republic and the Ministry of Transport of the Czech Republic under the Transport 2020+ Programme.
Michal Kvasnička is an assistant professor at the Department of Economics at FEA MU. His research interests are mainly in empirical microeconomics and data analysis. He teaches the full-faculty course Microeconomics I as well as the courses Analysis and Visualization of Economic Data and Applied Identification Strategies.
With the end of the year approaching, we bring you a summary of the most interesting events and achievements in the field of science and research at ECON MUNI. Our scientists have been involved in projects with a direct impact on society, celebrated success in grant competitions and helped politicians and public administration to make better decisions through the new Policy Reports Series.
The Dean's Award for Outstanding Dissertation Thesis was awarded in September to Katarína Čellárová, a graduate of doctoral studies at ECON MUNI and an experimental economist. In her doctoral research she found, among other things, that exploitation is less common when people are aware that they may switch roles with the victim in the future. It also revealed that winners who sacrificed more to win felt they deserved a better reward.
