Abstract An accurate prediction of incident duration plays an important role in obtaining traffic information for travelers timely and making appropriate decisions for traffic managers. As the characteristics are significantly different from each other, models were established for each type of incident, i.e. stopped-vehicle incidents, lost-load incidents and accidents. After data pretreatment, Principal component analysis (PCA) was carried out for each type of incident. Afterward, Naive Bayes (NB) model was applied for the data processed after PCA to predict incident durations. The experimental results indicated that the model obtained high prediction accuracy for those incidents which lasted less than 60 min and the prediction performance of accidents worked best. Besides, the prediction accuracy was 77.46%, 82.08% and 86.34% for each type of incident within 20 min’ error, respectively. In conclusion, the results showed that the combined model of PCA and NB is a promising application to predict incident duration.
Combined PCA and NB to Predict Traffic Incident Duration
International Symposium for Intelligent Transportation and Smart City (ITASC) 2017 Proceedings ; 1-11
2017-01-01
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Modeling Traffic Incident Duration Using Quantile Regression
Transportation Research Record | 2016
|Hazard based models for freeway traffic incident duration
Elsevier | 2012
|Hazard based models for freeway traffic incident duration
Online Contents | 2013
|Competing risks mixture model for traffic incident duration prediction
Elsevier | 2014
|Application of nonparametric regression in predicting traffic incident duration
DOAJ | 2018
|