详情

原文传递 Mining of Archived ITS Data Using Advanced Statistical Quality Control Techniques

题名：	Mining of Archived ITS Data Using Advanced Statistical Quality Control Techniques
作者：	Rod E. Turochy; Brian L. Smith, Ph.D.
关键词：	Mining; Archived; ITS; Advanced Statistical Quality Control Techniques
摘要：	The collection of data describing traffic conditions is a primary function of a typical traffic management system. Measurements of traffic parameters such as mean speed, volume, and occupancy are made at many locations on freeways and arterials covered by these systems. Spatially, traffic detectors are typically placed at regular intervals on freeways within these systems (e.g. every 0.5 mile). Temporally, these data are usually collected and aggregated across consistent time intervals (between 20 seconds and 5 minutes). Archiving of these data, by the provision of long-term storage in an easily accessible database, is becoming increasingly common in the upgrading and expansion of traffic management systems. As data storage technologies have advanced rapidly in recent years, so has an interest in tools to “mine” these databases and derive useful information from them. In the case of archived traffic data, one application of data mining is generating information describing patterns of abnormal traffic conditions. Knowledge of the history of abnormal traffic conditions and their relationship to temporal and spatial variables can be a great asset in the operation of traffic management systems. Anticipation of abnormal conditions based on historical data enables traffic managers to operate their systems in a proactive manner through pre-emptive activation of traffic control devices and dissemination of information to travelers. A data mining application to analyze archived data and thereby derive useful information of abnormal traffic conditions is under development at the Smart Travel Laboratory at the University of Virginia using archived data from the Hampton Roads Smart Traffic Center in southeastern Virginia; however, the methods described herein are easily transferable to archived data from any traffic management system. Since the data mining application operates on archived data, it can operate in an “offline” manner and does not require access to current “real-time” data. Additionally, the application need not run continuously but only periodically as enough new data are archived as to warrant an update of the resulting analysis or as any special analysis of traffic patterns is required. The key component of this data mining application is the use of statistical quality control techniques to detect abnormal observations. These statistical tools, developed to monitor manufacturing processes, examine quantified measures of such a process, identify observations that fall outside of a range defined as comprising normal conditions, and allow the process manager to take appropriate remedial action. These tools can be applied to a database of historical observations of traffic flow to perform the same function. Three measures of traffic flow typically collected in traffic management systems are mean speed, volume (flow rate), and occupancy. Traffic conditions can be considered normal when the values of the observed traffic variables fall within ranges that can be defined by historical data. The greater the deviation of a current observation from its expected value, based on historical data, the extent to which current traffic conditions are abnormal increases as well. When the system is not operating normally, or as expected, the abnormal conditions that are occurring can be detected through statistical quality control. Since mean speed, volume, and occupancy are related measures, a statistical tool that considers the interdependence of these measures is most appropriate. In this manner, multivariate statistical quality control treats the process measures as a set in its determination of the extent to which current traffic conditions are abnormal.
总页数：	ITS America. Meeting (11th : 2001 : Miami Beach, Fla.). ITS connecting the Americas : ITS 2001 : conference proceedings. 2001. pp13
报告类型：	科技报告