Accurately inferring the spatiotemporal distribution of freeway traffic volume is one of the bottleneck problems for intelligent management of ground transportation. Although the accuracy of the widely used machine learning-based approaches has been improving, these methods on freeway traffic flow inference tend to become more and more complex and unexplainable, and sometimes even contradict the constraints of freeway access. From the perspective of the potential destination city attractiveness, this study proposes a new method to infer freeway truck volume. First, considering the spatial scope and direction constraints of road transportation, a restricted area search method for potential destination cities of freeway traffic trips is proposed. Second, a freeway traffic inference model was constructed by combining distance and the angle-weighted socioeconomic statistics of each potential destination city. Finally, the Shapley additive explanation value was used to explain how the potential destination city attraction affects the freeway traffic flow. The results show that the proposed method only used 10 indicators; however, the prediction results were closer to the true value, and the computational efficiency improved by 3–10 times over the baseline method. The variables related to spatiotemporal heterogeneity were the most significant variable group, followed by the industrial structure of potential destination cities.
Read full abstract