Traffic Forecasting for King Fahd Causeway Using Artificial Neural Networks

Traffic prediction involves forecasting traffic in terms of Annual Average Daily Traffic (AADT), Design Hour Volumes (DHV) and Directional Design Hour Volumes (DDHV). These forecasts are used for a wide variety of purposes from the planning to the design and operational stages of the highway network. The forecasting needs the sufficient historical traffic data to establish the trend of traffic demand. Apart from that choice of an appropriate model or technique is also an important consideration that can affect the accuracy of results. Artificial Neural Networks are considered to be more suitable than traditional techniques like linear regression or moving average method in terms of their accuracy, flexibility and implementation in real-world situations. In this paper, we have developed two neural networks to predict the daily traffic flow on King Fahd causeway. Three years of data has been used to train the networks, initial results were flawed, thus to improve the performance the data was classified. The neural networks were then trained with the classified data which showed enhanced performance. Neural networks were trained with various parameter settings and the best arrangement was selected. Ideas regarding future work and enhancements are also presented.