Traffic Incident Duration Prediction Based On Partial Least Squares Regression

The prediction of the traffic incident duration is a very important issue to the Advanced Traffic Incident Management (ATIM). An accurate prediction of incident duration makes a lot contributes to making appropriate decisions to deal with incidents for traffic managers. The paper employed the Partial Least Squares Regression (PLSR) to build model between incident duration and its influence factors. Three models were established for three types of incident correspondingly, i.e. stopped vehicle, lost load and accident. Meanwhile, a model without distinguishing the incident type was built as a comparison. The experiments results indicated that the model obtained high prediction accuracy for those incidents which last 20 minutes to 90 minutes. The models got prediction accuracy of 77.24%, 86.59 %, 83.33% and 71.30% for stopped vehicle, lost load, accident and all incidents within 20 minutes error, respectively. The results indicated that the PLSR has a promising application to predict traffic incident duration (C) 2013 The Authors. Published by Elsevier Ltd. Selection and peer-review under responsibility of Chinese Overseas Transportation Association (COTA).