Advanced traffic control systems, such as traffic signal systems, include large numbers of sensors intended to support the monitoring of traffic conditions. In addition, transportation agencies frequently archive data collected by these detectors, on the assumption that important information can be extracted from the archives with the proper tools. The development of a data mining tool intended to support the maintenance of traffic signal systems that operate in the time-of-day (TOD) mode by identifying when traffic conditions have changed significantly in a corridor is described. The data mining approach used is classification. A case study was conducted to demonstrate that accurate classification models can be developed by using archived data to map between a set of traffic conditions and the associated TOD interval or timing plan for which the conditions are best suited. The 92.4% classification rate achieved in the case study indicates that this data mining tool has the potential to effectively support TOD signal operations.
Abstract