Analyzing the performance of genetically designed short-term traffic prediction models based on road types and functional classes

Abstract

Previous research indicates that the accuracy of short-term traffic prediction has close relationship with road type, and different models should be used for different types of roads. Road types are characterized by trip purposes and trip lengths. However, research for evaluating short-term model's performance based on functional classes is largely absent. Road's functional class is based on the functionality of individual road in a highway network. In this study, genetically designed locally weighed regression (LWR) and time delay neural network (TDNN) models were used to predict short-term traffic for eight rural roads in Alberta, Canada. These roads belong to different functional classes from four trip pattern groups. The influence of functional classes on the performance of short-term prediction models were studied for each type of roads. The results indicate that not only road trip pattern groups but also functional classes have large influence on the accuracy of short-term traffic prediction.