Short-term traffic volume prediction using neural networks (original) (raw)

There are many modeling techniques that can predict the behavior of complex systems, such as traffic volumes in regional transportation systems, with high accuracy. However, predictive power suffers significantly when non-recurring events, such as adverse weather, occur in these systems. Therefore, introducing novel ways to identify and quantify disruptions can improve projection accuracy and performance. Proactive traffic management requires the ability to predict traffic conditions. A relatively new mathematical model, the neural network, offers an attractive approach to modeling undefined, complex, and nonlinear situations. This algorithm is trained by using both historical data and non-recurring phenomena such as weather. In this study, we test our algorithm on traffic data collected on four highways and high-resolution weather data within the Dallas area. The test indicates the model’s high accuracy and efficiency in predicting short-term