Institute of Transportation, MOTC

       Predicting the traffic conditions is an important part of ITS. The purpose of this research is to develop a model that is able to integrate multiple detector and predict traffic conditions on highway. The data source of the proposed model is from VD and ETC data. This study fuses them as input of convolutional neural network (CNN) to predict traffic conditions.
       This study uses two methods to fuse ETC and VD data and transforms them into space-time matrices, and then uses these matrices as input of convolutional neural network to predict traffic conditions. The empirical testing results show that both single data and merging data which average MAE are below 6. Merging data perform well than single data in long-term prediction. It shows that the integration of multiple data sources and the CNN model can make accurate prediction on highway.