Researchers from the University of Zurich describe the results of urban systems (prediction of the car accident risk of individuals based on trajectory, driving events and geographical context): urban research

2022 APR 06 (NewsRx) — By a News Reporter – Staff News Editor at Daily Ecology News — The researchers detail the new data in Urban researchUrban systems. According to reports from Zürich, Switzerland, by NewsRx reporters, the research said, “With the prevalence of GPS tracking technologies, car insurance companies have begun to adopt usage-based insurance policies, which tailor insurance premiums based on customer driving behavior. Although there are many risk models to assess an individual driver’s crash risk based on historical driving trajectories, driving events and exposure records, these models do not hold account of the geographical context of driving trajectories and driving events.

Financial support for this research came from European Commission.

News correspondents got a quote from the research of the University of Zurich“This study explores the influence of enriching the existing purely driving behavior-based feature set with multiple geographic context features for the task of differentiating between accident and accident-free drivers. of five machine learning classifiers – logistic regression, random forest, XGBoost, feedforward neural networks (FFNN) and long-short-term memory networks (LSTM) – were evaluated on the usage records of more than 8000 vehicles in one year from Italy. The results show that the inclusion of geographical information such as weather, points of interest (POI) and land use can increase the relative predictive performance in terms of AUC by up to 8%, among which the land use is the most informative. For the data in this study, XGBoost generally gave the best performance and used geographic information the most, while logistic regression is only slightly outperformed by more complex models if the offered geographic information is not available. LSTM did not outperform other methods, perhaps due to the small amount of training data available.

According to the reporters, the research concluded: “The results highlight the potential of including geographic context in usage-based auto insurance risk modeling to improve accuracy, leading to location-based insurance policies. more equitable use.”

This research has been peer reviewed.

For more information on this research, see: Predicting Individuals’ Car Accident Risk By Trajectory, Driving Events, and Geographical Context. Computer Environment and Urban Systems2022;93. Computer Environment and Urban Systems can be contacted at: Elsevier Sci LtdThe boulevard, Langford AlleyKidlington, Oxford OX5 1 GB, Oxson, England.

Our journalists inform that additional information can be obtained by contacting Cheng Fu, University of Zurich, Department of Geography, Zürich, Switzerland. Other authors of this research include Livio Bruehwiler, Robert Weibel, Haosheng Huang and Leonard Longhi.

The direct object identifier (DOI) for this additional information is: https://doi.org/10.1016/j.compenvurbsys.2022.101760. This DOI is a link to a free or paid online electronic document, and can be your direct source for a journal article and its citation.

(Our reports provide factual information on research and discoveries from around the world.)

Comments are closed.