In general, incident management and information dissemination strategies will benefit from the prediction of incident durations in real time. This study investigates the development of an incident duration prediction algorithm based on a detailed historical incident management database. A total of 2,629 incidents extracted by a data filter process were used in this study to predict the incident durations. A data mining technique, namely the Bayesian Network (BN) method is applied to develop incident duration prediction models. Based on the sequence of the incident duration process, two models were developed. The prediction accuracy of BN model for stage one is 45.7588%, while BN model for stage two has much higher prediction accuracy 72.5751 %. When compared with other data mining models, the goodness-of-fit results suggest that the BN model is advantageous in terms of higher prediction accuracy and the convenience of application.
Freeway incident duration prediction using Bayesian network
2017-08-01
258614 byte
Conference paper
Electronic Resource
English
Predicting Freeway Incident Duration Using Machine Learning
Springer Verlag | 2020
|Effective Freeway Incident Response: A Bayesian Network Based Algorithm
British Library Conference Proceedings | 2008
|