Double-Sided Determination of Aggregation Level for Intelligent Transportation System Data

Abstract

Intelligent transportation system (ITS) data, which are normally collected at a 15- to 30-s interval, are a rich and valuable resource for a variety of applications, including transportation planning. However, raw ITS data exhibit a wide range of fluctuations, which may not be directly useful for planning purposes, for which aggregations taken over longer time intervals are commonly needed. Proper determination of aggregation level of ITS data will ensure the retention of necessary information and the elimination of as much unnecessary information as possible. The traditional approaches to determining aggregation level are intuitive and easy to implement, yet they may be unable to determine whether particular information is kept or lost. The newly developed wavelet-based approach, although good for decomposing original data sets, needs to be further improved. A double-sided technique was developed that could refine the selection of aggregation levels based on two self-defined indices: the dissimilarity index and the information loss index. A computer program coded in MATLAB with the use of the Wavelet Toolbox was developed to implement the proposed technique. A case study illustrating the process was conducted by using real-time ITS data from traffic management centers in the San Antonio TransGuide. It is expected that the proposed technique will maximize the use of real-time ITS data and improve data needs in the urban transportation planning process.