Abstract
The paper reports on the evaluation of the performance of various short-term congestion prediction methods, i.e. multi-linear regression, time series analysis, multi-layer perceptions, radial basis function networks, self-organising systems, and fuzzy logic. Data were gathered through dual induction loops on an A10 motorway section in the Netherlands during a four-week period. The data consist of 1 minute aggregated time bins of volume, occupancy, speed, and both a reliability and a congestion indicator. The method's results are similar, except for multi-linear regression. Self-organising systems were omitted due to huge error production.