Travel-Time is one of the most important parameters for Intelligent Transportation Systems (ITS), Advanced Traveler Information Systems (ATIS), and Advanced Traffic Management Systems (ATMS). Real-time vehicle speed data is measured based on loop detector in ITS studies. Accurate traffic speed data is the raw element of the travel time estimation and calculation. Real-time traffic data from loop detectors are inevitably corrupted by unexpected missing values or appear to be given nonsensical or erroneous data due to detector faults or transmission distortion. Missing data handling is an important preparation step for data mining tasks in travel time estimation. Accurate traffic speed estimation can improve the quality of estimated travel time information for ITS.
In this study we present a novel algorithm to estimate continuous-time traffic speed data using Exponential Window based on the correlated speed and then compare its results to other baseline missing speed estimation methods with real freeway traffic speed data. Since this approach has greater generalization ability for given real speed data, it is believed that this model will also perform well for all time-series missing data estimation fields.
