There’s no doubt the number of deaths and injuries from road traffic accidents is decreasing. Back in 1996, there were just over 60,000 fatalities on European roads, compared to 25,100 in 20181, a decrease of nearly 60%. Safer roads, safer cars, better infrastructure and progressed laws have all played their part in improving our road safety. But how can we lower these numbers even more? With significant advancements in machine learning and increased availability of open data of high quality, the use of data is steadily increasing amongst public sector organizations in their aim to tackle risk management in public domains.
In 2018, BearingPoint collaborated with the Norwegian Public Roads Administration which is recognized for world-class road safety initiatives. The number of deaths from road accidents in Norway is very low with 108 deaths in 20182. But the organization wanted to go further in implementing preventative measures to minimize the risks of accidents on roads.
To find out what those preventative measures could be, we applied Hypercube - a machine-learning algorithm built to examine high-dimensional data in an exhaustive and transparent way, to gain understanding of complex risk factors. From the data provided, Hypercube offers accessible and interpretable results that contribute to a series of actionable insights. However, using unique algorithms doesn’t have to be your starting point to exploring new forms of risk management. Here is an approach to how you can start using data and analytics:
Data and analytics can be daunting, but it doesn’t have to be. Build a structured approach and expand your plans from there. You don’t always need complex IT integrations or a lot of software to create valuable insight from data. The building blocks to our work come from open data that is publicly available. When researching road traffic accident prevention, areas such as road construction, demographics, climate data, and a recent history of local road fatalities and injuries is all publicly available data. From there, it’s possible to carry out a one-off analysis to assess the potential avenues for further research and development. These initial stages can be quite short term, low cost solutions. You can get started with very little.
Make sure you identify a specific area in which you want to build your insights. Only then can you use data to truly understand a problem. A classic mistake that organizations often make is to assume that business benefits can automatically arise from large amounts of data. Quantity doesn’t mean much if you don’t have a target in mind. Look towards your organization first rather than the data. What are the most pressing business issues right now? Where do you need to improve? Produce a specific problem statement that data and analytics could support. For example, the basis for the problem statement of our work in Norway was: ‘What characteristics of the road and surroundings cause high rates of deaths and serious injuries?’ The more specific the problem statement is, the easier it will be to utilize the data in a targeted way, and the more useful the insights will be.
Once you’ve identified the specific area you want to focus on, and collected the data with the most potential, you’ll need to process your data and turn it into something that is prescriptive for the problem being analyzed. Data processing is often the most time-consuming part of the process. With a sensibly processed data set, however, algorithms can be effective in building actionable insights. For our case, identifying specific combinations of road construction elements, demographics and climates that run a high risk of accidents. These insights can be highly valuable in improving existing roads and designing new ones, potentially also taking local aspects into account.
Every day, data is expanding and becoming more and more detailed. And at the same time, the algorithms and the hardware built to analyze that data are becoming more powerful. Organizations that are able to harness these two factors using a smart, structured approach, who are able to produce specific problem statements tailored for business implementation, will put themselves in a stronger position on the journey towards data-driven insights. For the Norwegian Public Roads Administration, that means minimizing risks on the roads and potentially saving lives.