Citation

Kweon Y, Sun B, Park BB. Transp. Res. Rec. 2021; 2675(10): 268-276.

Copyright

(Copyright © 2021, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.1177/03611981211011162

PMID

unavailable

Abstract

While big data helps improve decision-making and model developments, it often runs into privacy concerns. An example would be retrieving drivers' origin and destination information from smartphone navigation apps for developing a route choice behavior model. To conserve privacy, yet to take advantage of big data in navigation applications, the authors propose to apply a federated learning approach, which has shown promising application in predicting smartphone keyboard's next word without sending text to the server. Additional benefits of using federated learning is to save on data communications, by sending model parameters instead of entire raw data, and to distribute the computational burden to each smartphone instead of to the main server. The results from real-world route navigation usage data from about 30,000 drivers over one year showed that the proposed federated learning approach was able to achieve very similar accuracy to the traditional centralized global model and yet assures privacy.

Language: en