Reproduce “Use ADAS Data to Predict Near-Miss Events: A Group-Based Zero-Inflated Poisson Approach” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xinbo Zhang, Montserrat Guillen, Lishuai Li, Xin Li, Youhua Frank Chen
Researchers from City University of Hong Kong and the University of Barcelona developed a framework using group-based zero-inflated generalized Poisson models to predict weekly near-miss events from multi-sensor ADAS and telematics data. This approach effectively addresses data sparsity, overdispersion, and driver heterogeneity, achieving superior predictive accuracy compared to traditional statistical methods.
Key insights
- Explicitly modeling both excess zeros and overdispersion with zero-inflated and generalized Poisson distributions significantly improves the statistical fit for near-miss event data.
- Grouping drivers into behaviorally homogeneous latent classes effectively captures individual heterogeneity, leading to more accurate predictions, especially for specific NME types like harsh acceleration.
- A comprehensive approach integrating multi-sensor ADAS warnings, kinematic telematics, and contextual data provides a robust foundation for proactive and dynamic driver risk assessment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.