Toward Improved and Transparent Imputation Techniques for Online Traffic Data Streams and Archiving Applications

A diverse range of measurements collected from the transportation infrastructure facilitate day to day operation, surveillance, forecasting, and dissemination of current condition information to the general public. The mechanisms for assessing current system conditions rely on multiple sensor types and mobile probes. The quality and completeness of traffic data is generally regarded as suboptimal. Several techniques are used in the transportation industry to cope with incomplete or suspiciously erroneous data, in particular the imputation of missing values in a range of types of traffic databases, streams and archives. The objective of this paper is to identify and categorize common imputation techniques reported in the transportation literature. This review will be discussed and presented in the context of a notional system that performs online imputation for traffic data streams. Exemplar uses of traffic data include traveler information systems and traffic management applications.