A categorical model for traffic incident likelihood estimation
MetadataShow full item record
In this thesis an incident prediction model is formulated and calibrated. The primary idea of the model developed is to correlate the expected number of crashes on any section of a freeway to a set of traffic stream characteristics, so that a reliable estimation of likelihood of crashes can be provided on a real-time basis. Traffic stream variables used as explanatory variables in this model are termed as Ã¢ÂÂincident precursorsÃ¢ÂÂ. The most promising incident precursors for the model formulation for this research were determined by reviewing past research. The statistical model employed is the categorical log-linear model with coefficient of speed variation and occupancy as the precursors. Peak-hour indicators and roadway-type indicators were additional categorical variables used in the model. The model was calibrated using historical loop detector data and crash reports, both of which were available from test beds in Austin, Texas. An examination of the calibrated model indicated that the model distinguished different levels of crash rate for different precursor values and hence could be a useful tool in estimating the likelihood of incidents for real-time freeway incident management systems.
Kuchangi, Shamanth (2006). A categorical model for traffic incident likelihood estimation. Master's thesis, Texas A&M University. Texas A&M University. Available electronically from