Data fusion algorithm improves travel time predictions

Travel time is considered more useful to users than other travel-related information such as speed. It is mainly estimated by point or interval detection systems. In this study, the authors investigate the deficiency of these systems in estimating travel times when they are used in isolation, and proposed a fusion algorithm that simultaneously utilises data from both point and interval detection systems. The fusion algorithm is based on the traffic flow and k-nearest neighbourhood (k-NN) models. Specifically, the authors precisely define the so-called the time lag issue in interval detection systems. To overcome this problem, they analysed the travel time variation because of variation in traffic states using fused data from point and interval detection systems. The authors show that the travel time obtained from interval detection systems is renewed by considering the travel time variation and their results show that the proposed algorithm satisfactorily predicts the travel time with the mean absolute percentage errors (MAPE).