From Time Based Speed Series to Behavior Sequence: The Transformations and Applications of VTDR Data
Zhenfeng He, Zhenzhi Huang · 2009
Vehicle Traveling Data Recorders (VTDRs) have been universally installed on fixed-route intercity buses, and the data collected is growing rapidly. It becomes urgent to utilize VTDR data to give drivers and managers some helpful feedbacks. Two novel representations of VTDR data: Location Based Speed Series (LBSS) and Behavior Sequence (BS) are presented for the analysis of these data. The data gathered directly from VTDR is Time Based Speed Series (TBSS), which defines the speeds at different time points. The fixed intercity route makes it feasible to covert a TBSS to an LBSS, which defines the speeds at different locations. BS is then proposed to extract the dynamic information, such as the speed change pattern at a segment of an LBSS. For converting an LBSS to a BS, a Background Based Symbolization (BBS) approach is presented. Based on the background of an average speed curve, five typical behavior patterns are defined. During the BBS process, each length-fixed segment of an LBSS is labeled as one of the five patterns, and the consequent pattern sequence is a BS. A simple example is presented to illustrate the transformations. The applications of LBSS and BS are illustrated with the analysis of FuQuan Highway data. The LBSS is effective in the evaluation of the driving behavior at Xiangsiling Tunnel. And BS is more suitable in mining sequential patterns than those symbolized by Symbolic Aggregate Approximation (SAX).