For multi-disciplinary studies of traffic and socio-economic analysis, it is essential to obtain a high-level summarization of the traffic condition at different regions of an urban area. A Traffic Index summarizes the traffic at the level of census blocks (or other zone types) and at different times of the day. In this work, we propose an algorithm that aggregates the traffic data of a large region and constructs a Traffic Index that is able to estimate the traffic at any given location in that region. Complete and accurate traffic flow data is required in order to accurately summarize the traffic. However, in reality, it is hard to acquire such data due to various reasons such as lack of enough sensors, hardware malfunctions, and software issues in traffic sensing. Our research center has been collecting sensor-based traffic data in the metropolitan city of Los Angeles for more than 10 years and our observation shows that data has been missing on a daily basis, resulting in sparse and incomplete traffic datasets. Therefore, we evaluate several imputation methods to recover such missing data and make the dataset complete. Additionally, in order to overcome the sparsity of traffic sensors, we evaluate spatial interpolation methods. We conduct comprehensive experiments to study how these imputation and interpolation methods perform. Lastly, we provide a unique Traffic Index case study in the metropolitan city of Los Angeles using our real-world traffic data.
Data-driven Traffic Index from Sparse and Incomplete Data
2022-10-08
1212549 byte
Conference paper
Electronic Resource
English
Data-Driven Evaluation Methodology for Active Traffic Management Systems Utilizing Sparse Speed Data
Transportation Research Record | 2023
|Path Travel Time Estimating Method by Incomplete Traffic Data
Springer Verlag | 2020
|