Cluster-based lognormal distribution model for accident duration

Publication Type:
Journal Article
Citation:
Transportmetrica A: Transport Science, 2015, 11 (4), pp. 345 - 363
Issue Date:
2015-01-01
Full metadata record
Files in This Item:
Filename Description Size
Cluster based lognormal distribution model for accident duration.pdfPublished Version652.3 kB
Adobe PDF
© 2015 Hong Kong Society for Transportation Studies Limited. This study develops a cluster-based lognormal distribution model for the purpose of predicting accident duration. With Maryland I-95 freeway traffic accident data collected in 2010 and 2011, this study first uses a decision tree approach to split the entire sample data into three clusters which are then treated as additional variables in modelling accident duration. The results show that seven explanatory variables and cluster variables significantly affect the mean accident duration. With the cluster-based lognormal distribution model, the mean and the probability of an accident duration being unacceptable can be predicted from the base accident information. Such predictions can be utilised as a basis for making rational diversion in the event of an accident, which will help mitigate traffic congestion and improve travel time reliability.
Please use this identifier to cite or link to this item: