Missing Data Treatment and Data Fusion toward Travel Time Estimation for Atis

This study develops a travel time estimation process by integrating a missing data treatment and data-fusion-based approaches. In missing data treatment, this study develops a grey time-series model and a grey-theory-based pseudo-nearest-neighbor method to recover, respectively, temporal and spatial missing values in traffic detector data sets. Both spatial and temporal patterns of traffic data are also considered in travel time data fusion. In travel time data fusion, this study presents a speed-based link travel time extrapolation model for analytical travel time estimation and further develops a recurrent neural network (RNN) integrated with grey models for real-time travel time estimation. In the case study, field data from the national freeway no.1 in Taiwan is used as a case study for testing the proposed models. Study results showed that the grey-theory-based missing data treatment models were accurate for recovering missing values. The grey-based RNN models were capable of accurately predicting travel times. Consequently, the results of this study indicated that the proposed missing data treatment and data fusion approaches can ensure the accuracy of travel time estimation with incomplete data sets, and are therefore suited to implementation for ATIS.