Abstract Given the reliance on passive safety control, this study proposes a real-time risk warning model. It integrates the feature selection of the random forest algorithm with the decision optimization of matter-element theory, analyzing 2023 traffic flow data and accident records from California’s PeMS system for I880-N. This identifies correlations among traffic flow factors leading to accidents, revealing patterns in frequent accident zones. An initial indicator system for highway risk states is constructed based on traffic flow, speed, and occupancy rate. Using the random forest method, 13 key warning features are identified and weighted. K-means clustering categorizes risk states into basic levels. Applied to high-accident sections of I880-N, the model aims to improve real-time hazard anticipation and mitigation, shifting from reactive to proactive safety management. The study shows increased variability in occupancy and speed with accidents, and significant traffic flow changes five minutes before accidents. The model’s fit for real-time risk assessment is 0.0056, indicating severe risk (Grade Four). This enhances danger identification, reduces traffic accident rates, and introduces a new paradigm for road risk early warning.
Advanced Real-time Highway Risk Early Warning System Integrating Comprehensive Weighting Techniques
Int. J. ITS Res.
12.06.2025
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Highway risk active early warning system integrating sound, light and electricity
Europäisches Patentamt | 2020
|Real-time tracking highway traffic safety early warning system
Europäisches Patentamt | 2023
|Real-Time Early Warning Techniques
NTRS | 2016
|Movable highway traffic weather real-time monitoring and early warning system
Europäisches Patentamt | 2024
|Highway agglomerate fog early warning system and early warning method
Europäisches Patentamt | 2023
|