A deep learning framework to generate realistic population and mobility data
This work addresses data scarcity and privacy issues in travel demand estimation and agent-based modeling, though it appears incremental as it builds on existing synthetic data augmentation methods.
The paper tackles the problem of generating realistic synthetic population and mobility data to address limitations in census and travel survey datasets, achieving competitive performance on multiple assessment metrics compared to other recent models.
Census and Household Travel Survey datasets are regularly collected from households and individuals and provide information on their daily travel behavior with demographic and economic characteristics. These datasets have important applications ranging from travel demand estimation to agent-based modeling. However, they often represent a limited sample of the population due to privacy concerns or are given aggregated. Synthetic data augmentation is a promising avenue in addressing these challenges. In this paper, we propose a framework to generate a synthetic population that includes both socioeconomic features (e.g., age, sex, industry) and trip chains (i.e., activity locations). Our model is tested and compared with other recently proposed models on multiple assessment metrics.