01031408

Component

Artificial Neural Network (ANN) is regarded as an excellent candidate to model complex traffic prediction problems. However, there is no guideline on the choice of the typologies of neural networks. Also, very few studies have been conducted on the topic of traffic prediction in the presence of traffic incidents due to many difficulties such as data source, data fusion and modeling problems. This research utilizes three different types of neural networks to model corridor travel time prediction in the presence of traffic incidents using data collected from a highway corridor in Northern Virginia. Sensitivity analysis is employed to test the relative importance of the different input channels, and the test results reveal that some input channels always play significant roles, while some others are always insignificant. The constantly insignificant input channels can be regarded as superfluous and be eliminated from the model to reduce the complexity of the training process and future data collection costs. The performances of the three different neural networks are compared, and the results demonstrate that under some cases, one typology of neural network performs better, and under other cases, another typology is superior. This finding indicates that different types of neural network may specialize in different portions of the input pattern space, thus it may be a promising area to develop new types of neural networks which can take advantage of the benefits of different types of networks in traffic prediction.

01020180

06-2512

English

24p

2006

Transportation Research Board 85th Annual Meeting

CD-ROM

Figures (8) ; References (8) ; Tables (2)

Data collection; Data fusion; Incident detection; Neural networks; Traffic; Traffic congestion; Traffic flow; Traffic forecasting; Traffic incidents; Travel time

Northern Virginia

Highways; Operations and Traffic Management; Planning and Forecasting; I72: Traffic and Transport Planning

Transportation Research Board Annual Meeting 2006 Paper #06-2512

TRIS, TRB

Mar 3 2006 11:02AM