- The Flemish government’s Department of Mobility and Public Works (MOW) actively supports the Ministry in its policy concerning mobility and traffic safety, as well as investment, management and operation of the transportation and port infrastructure in Flanders.
- The Flemish Minister for Mobility implemented a Traffic Safety Plan to reduce the number of traffic fatalities to zero by 2050. Education, enforcement, engineering and research are all part of the plan.
- MOW has access to an enormous amount of traffic information, comprising multiple types and structures, and coming from many sources including police databases (accident data), weather history, road conditions, and more. Much data also comes from magnetic detection loops, which are electronically controlled wires installed across the road surface at hundreds of locations throughout Flanders’ highway system. The magnetic loops detect and record basic information about every vehicle that runs over the wire. The loops register the location and time of the passage, the length of the vehicle, its speed and the elapsed time since the last passage.
- In partnership with universities, MOW did a lot of academic research on the data coming from magnetic loops, to be able to use it for mobility purposes. They now want to take that knowledge to the next level.
- MOW does not have an enterprise analytics platform, nor the methodologies or processes capable of supporting rapid data discovery and complex analysis of large volumes of structured and unstructured data.
- MOW approached DXC Technology to help them find a way to use all available data to fuel data-driven decision-making.
- DXC Technology suggested a Big Data Discovery Experience, including a one-day workshop with MOW business owners to decide on project scope, and eight weeks of thorough data analytics.
- During the workshop, participants identified 30 relevant use cases. They also aligned objectives, listed all available data sources, and agreed on task owners and next steps.
- After a prioritization of the use cases, one stood out: Use the huge amount of raw data, coming from magnetic traffic detection loops to recognize and analyze new factors that lead to accidents. Two major questions were put forward as input for the data analytics exercise: “Why do accidents happen?” and “Can we define safe versus unsafe roads?”
- The workshop was followed by an eight-week period of data acquisition, integration and advanced analytics. DXC Technology industry-experienced consultants and data scientists with advanced analytics skills worked with MOW to capture the value of their data – supported by the necessary tools and a proven methodology.
- For this particular exercise, the project team received around 10 percent of the available data, comprising more than 6 billion records. The data scientists sliced and aggregated the raw information across multiple dimensions (time, location, etc.), and combined it with accident information received from the police as well as various other data sets that were deemed relevant for accident risk predictions (road quality, weather conditions, sunlight conditions, etc.).
- The analysis resulted in a large, extensive data set that contained hundreds of traffic indicators providing a clear view on traffic dynamics, both looking at individual locations as well as complete road segments.
- Various clustering techniques were applied to separate safe from risky locations, based on a first dataset of the more dangerous road segments. The team also investigated which traffic variables would increase the likelihood of accidents using an iterative approach of modelling, validation and fine-tuning.
- DXC Technology then built several predictive models based on density and intensity patterns that were presented to MOW at the end of the eight-week period. The models make it possible to define accidents risk factors for a certain location and at particular point in time. The model was validated in depth on the Antwerp ring and was applied on several other road segments, illustrating that the used modelling techniques can be generalized.
Value for MOW
- The MOW Big Data Discovery Experience demonstrated that an enormous amount of data can indeed be used to determine accident factors and derive an overall accident risk. The insights that MOW gained from the predictive models are now under consideration for further deployment as way to address the 2050 objective of zero traffic fatalities.
- MOW took the opportunity to fully learn from the data discovery exercise. They were active participants in the entire process, building experience and challenging the way they handle business issues today. It was a truly collaborative and mutually inspiring journey.
- As the project matured, MOW began to envision different ways to deploy the models. For example, the traffic control center could use safety indicators to draw more attention to certain road segments. Another idea was to use the model predictions to evaluate the safety impact of certain changes to the actual road infrastructure. These ideas and other will be evaluated by the various stakeholders