Generating mobility networks with generative adversarial networks

Abstract

The increasingly crucial role of human displacements in complex societal phenomena, such as traffic congestion, segregation, and the diffusion of epidemics, is attracting the interest of scientists from several disciplines. In this article, we address mobility network generation, i.e., generating a city’s entire mobility network, a weighted directed graph in which nodes are geographic locations and weighted edges represent people’s movements between those locations, thus describing the entire mobility set flows within a city. Our solution is MoGAN, a model based on Generative Adversarial Networks (GANs) to generate realistic mobility networks. We conduct extensive experiments on public datasets of bike and taxi rides to show that MoGAN outperforms the classical Gravity and Radiation models regarding the realism of the generated networks. Our model can be used for data augmentation and performing simulations and what-if analysis.

Mobility constraints in segregation models

“…In addition, there is potential for exploring alternative mechanisms of agent mobility. For instance, to capture individual mobility patterns more realistically, we could assign individual mobility networks to agents 56 , 74 , limiting their movement to specific subsets of cells on the grid. Lastly, we aim to extend the analysis to a real-world dataset of relocations to assess how simulation-based results align with empirical observations in an actual city.…”

A novel approach to model the role of mobility suppression and vaccinations in containing epidemics in a network of cities

“…The study also clearly shows the significance of maintaining control strategies while vaccinating the population. Mobility between regions is frequently generated using mobility networks with weighted directed graphs where nodes represent geographical locations and edges represent mobility flows between locations ( Ganciu, Balestrieri, Imbroglini, & Toppetti, 2018 ; Martin, Wiedemann, Reck, & Raubal, 2023 ; Mauro, Luca, Longa, Lepri, & Pappalardo, 2022 ). The topology of such networks has a significant impact on the epidemic's spread ( Moreno, Pastor-Satorras, & Vespignani, 2002 ).…”

CATS: Conditional Adversarial Trajectory Synthesis for privacy-preserving trajectory data publication using deep learning approaches

“…More importantly, limited attention has been paid to the privacy protection capability in existing deep learning approaches. Although recent works have applied deep neural networks (e.g., generative adversarial network, GAN) in synthetic trajectory data generation such as LSTM-TrajGAN (Rao et al, 2020), TrajGAIL (Choi et al, 2021), TrajGen (Cao and Li, 2021), Deep Gravity (Simini et al, 2021), and MoGAN (Giovanni et al, 2022), the key differences between this work and existing ones are as follows: 1) we provide a distributional-level K-anonymity privacy guarantee and verify the privacy protection effectiveness by experiments, which was not investigated in TrajGAIL, TrajGen, Deep Gravity, or MoGAN. LSTM-TrajGAN also examines its privacy protection effectiveness by experiments, but it did not consider the conditional spatiotemporal data distribution; 2) we achieve conditional trajectory generation by using aggregated human mobility distribution as a condition and model trajectory global context using the attention-based mechanism, which was not supported in other works; and 3) we generate individual-level synthetic trajectory data to preserve spatiotemporal mobility patterns of raw trajectory data at a city scale.…”