Optimizing Traffic Prediction Performance of Neural Networks under Various Topological, Input, and Traffic Condition Settings

Abstract

This paper presents an approach to optimize the short-term traffic prediction performance on freeways using multiple artificial neural network topologies under different network and traffic condition settings. The approach encourages multimodel techniques that are capable of improving the prediction system performance over single-model approaches. Using a mix of neural network topologies, the short-term speed prediction performance was extensively evaluated under different input settings and various prediction horizons (from 5 to 20 min). To enable the networks to learn from historical information, a long-term memory component was introduced to the input patterns to allow the networks to build internal representation of recurrent conditions, in addition to the short-term memory that is encoded in the most recent information. Optimal settings were determined by maximizing the performance under different traffic conditions observed at the target location, as well as upstream and downstream locations. Comparative statistical analysis with naive and heuristic approaches showed that the optimized neural network approach resulted in better prediction performance. The study shows that the optimal settings were consistently more dependent on the long-term memory component as prediction horizon increases.