The research that is reported describes two critical steps in developing a methodology for extracting secondary accidents from police accident databases. The eventual goal of the research is to produce an incident progression model from which the number of secondary incidents could be determined. These steps include the following: (1) the development of models for incomplete incident data and (2) an analysis of site differences in maximum queue lengths and incident duration. The first step is important because it allows the use of intranet incident data that is complementary to police accident data. The second step is also important because it helps determine whether site-specific models are necessary instead of a single general model. The methodology involves the use of data fusion from accident data and intranet traffic reports. The project results in a near-term technology for analyzing the safety impacts of transportation assets.
Abstract