Identifying Users Transferring Between Transportation Modes: A Stable Matching Approach

Abstract

Travel data containing personal identification information needs to be anonymized before being shared and analyzed for privacy considerations. While this approach protects personal privacy, it makes it difficult for researchers and planners to identify the same traveler from different databases and to construct complete multi-modal trips, which greatly reduces the value of data. To address this challenge, this paper develops TBLink, a method to match individual travelers between different modes. The underlying idea is that if a traveler makes a transfer, the spatiotemporal signatures of the previous and next trips will be similar and must satisfy certain conditions. When this pattern occurs for two travelers in the two datasets repeatedly, we can infer that the two travelers are actually the same person. The matching of travelers is regarded as a stable matching problem, and the Gale-Shapley algorithm is used to solve the problem. TBLink is demonstrated using metro and bus trip data from Chengdu City between January and March 2021. The results show that the precision, recall, and recovery rate of user matching are 80.83%, 92.82%, and 10.56% respectively, and the matching is more reliable as the dataset increases. Sensitivity analysis is performed to study the effect of several model parameters on the matching performance.