Recent developments in GIS and analytical applications have demonstrated that predicting road conditions, and thus preventing traffic accidents and possibly even traffic in the first place, could be possible. Among other goals, Vision Zero is an initiative by cities and municipalities to reduce traffic fatalities to zero. GIS and spatial analysis could help achieve this goal.
One tool, called Hopper, has been developed by the tech startup Open Data Nation. This is a machine learning tool that takes previous data, such as accidents and conditions on a given day, and then predicts the likelihood similar events may occur in a given area. As conditions might be comparable to previous events, these predictions could then be based on historical patterns that inform drivers so that they could avoid certain areas. For example, some intersections may be more likely to cause accidents during wet or wintry conditions. New developments in this tool can allow it to be integrated with ArcGIS and is sponsored by major technology companies such as Microsoft. The tool even purports to find restaurants that are likely to fail health safety checks based on historical patterns.
The Hopper tool is part of a larger trend in GIS analytics that has seen dynamic mapping tools developed which account for real-time data and predictive software that can inform users likely places where events, such as accidents, are likely to occur. Smartphones using cloud-based services can retrieve live data from GPS devices from users, integrating historical data with known weather conditions, including what happened in the past when similar conditions were present. Machine learning methods, including artificial neural network techniques, allow a measure of confidence to be built through comparisons and probability matching current conditions with the past.
While applications such as Hopper utilize data from previous road conditions, other approaches directly look at the drivers themselves. Sensors that pickup driving behavior, including those used with smartphones, allow a detection of speed relative to speedlimit, style of driving, and other behaviors that could be classified. This classification could allow a larger profile set to be built for many drivers, where the results could be interpolated for the larger population. Predictions then could be based on the demographic of driver profiles relative to given conditions. Thus, if a certain percentage of drivers, who might be classified as relatively reckless, were to drive in a day where conditions were less than optimal, then road accidents could also be predicted based on likely human behavior and not just general conditions. This is the possibility now available by combining on-board diagnostic adapters with GPS data gathering.
The likely trend of these developments is that as vehicles begin to increasingly be driver-less or at least provide driver assistance, these technologies of prediction could be directly incorporated through cloud-based services that transfer data directly to one’s car. Only a few seconds before a specific area is encountered would be needed for cars to then maneuver as needed based on predictive rather than real-time data. In fact, cars would be automatically informed to avoid given areas if conditions proved difficult. These cars can incorporate video, sensor, and wider data from other vehicles using cloud services. Currently this has been applied for driver assisted vehicles but there is no reason why they will not be extended to driver-less vehicles.
What we see is a trend of where tools now allow drivers to predict routes based on known conditions at a given time. Most technologies utilize historical data collected which then look for patterns and utilize classification, such as through random forest techniques, that then allow machine learning algorithms gauge how similar current conditions are. Sensors and cameras have been utilized along with the wider road networks such as through GPS data. Driver behavior is also critical and sensors that monitor this have been of greater interest, particularly for the insurance industry. In the near future, however, monitoring driver behavior maybe less needed, as future applications are likely to see driver-less vehicles apply predictive technologies to road conditions.
 For more on comparable tools to Hopper and methods used to apply machine learning techniques for road condition predictions, see: Gawad, S. M. A., El Mougy, A., & El-Meligy, M. A. (2016). Dynamic Mapping of Road Conditions Using Smartphone Sensors and Machine Learning Techniques (pp. 1–5). IEEE. https://doi.org/10.1109/VTCFall.2016.7880972.
 For more on driver profiling and how that is used to classify and, thus, allow prediction of driving, see: Ferreira, J., Carvalho, E., Ferreira, B. V., de Souza, C., Suhara, Y., Pentland, A., & Pessin, G. (2017). Driver behavior profiling: An investigation with different smartphone sensors and machine learning. PLOS ONE, 12(4), e0174959. https://doi.org/10.1371/journal.pone.0174959.
 For more on how driver assisted vehicles are using new predictive technologies for road conditions, see: http://openaccess.thecvf.com/content_iccv_2015/papers/Jain_Car_That_Knows_ICCV_2015_paper.pdf.
- OpenRouteService’s “Quiet Routing” Creates a Less Stressful Route for Pedestrians
- A Spatial Analysis of Moscow’s Fuel Stations
- Mapping Noise Pollution