Abstract: A scheme to forecast shortterm traffic volumes of a road network is presented. This scheme combines the principal component analysis and the support vector regression. First, the traffic volume data of multiroadcrosssections acquired from a road network are turned into several time series data by the principal component analysis. Then, these principal component data are used to train the support vector machines, and a genetic algorithm is applied to optimize the parameters of the support vector machines. After the training and optimization, by inputting required data to the support vector machines, the principal components are obtained as the outputs. These outputs are then transformed into the forecasting data of the shortterm traffic volumes. A case study is carried out to validate the scheme. The traffic volume data from seven road crosssections of an urban expressway are used. The forecasting results by this scheme are much better than that by the schemes of single road crosssection.

Key words: intelligent transportation systems, short-term traffic volume forecasting, support vector machine, principal component analysis, road network