Abstract Mobility models have a broad range of applications in areas related to human movements, such as urban planning, transportation, and simulations of diseases spread. In the last decade, the extensive geolocated user trajectories collected from mobile devices allowed for more realistic mobility modelling, improving its accuracy. However, mobility data sharing raises privacy concerns, which in turn limits accessibility to the data. In this paper, we propose a WHO-WHERE-WHEN (3W) model, an improved privacy-protective mobility modelling method for synthetic mobility data generation. Based on real trajectories, it produces artificial user mobility trajectories that simulate population fluctuations in a study area, and thus preserves the individual's privacy. The model simulates the individual spatiotemporal aspects of lives accurately, representing real population flows and distributions. The proposed method was inspired by the Work and Home Extracted REgions (WHERE) algorithm, but we have extended it by considering the activity space and circadian rhythm of people. Furthermore, we propose a clustering approach to capture and reproduce the heterogeneous characteristic of mobility. We evaluate our model and compare its performance to the WHERE algorithm on the synthetic and real data test cases. Use of the 3W model improved the accuracy of population distribution reproduction by 35% measured using Earth Mover's Distance. The travel distances and the spatial distribution of the flows reproduced by the 3W model match input data with high accuracy. We also evaluate the level of privacy protection by comparing synthesised and input datasets. We find that no daily trajectory can be matched between input and synthesised datasets and the average length of the matching sequence of visited locations to contain only two locations.

中文翻译：

---

用于流动数据模拟的人口流动建模

摘要 移动性模型在与人类运动相关的领域有广泛的应用，例如城市规划、交通和疾病传播模拟。在过去十年中，从移动设备收集的广泛地理定位用户轨迹允许更逼真的移动建模，提高其准确性。然而，移动数据共享引发了隐私问题，进而限制了数据的可访问性。在本文中，我们提出了 WHO-WHERE-WHEN (3W) 模型，这是一种改进的隐私保护移动建模方法，用于合成移动数据生成。它基于真实轨迹，生成模拟研究区域人口波动的人工用户移动轨迹，从而保护个人隐私。该模型准确地模拟了生活的个体时空方面，代表真实的人口流动和分布。所提出的方法受到工作和家庭提取区域（WHERE）算法的启发，但我们通过考虑人们的活动空间和昼夜节律对其进行了扩展。此外，我们提出了一种聚类方法来捕获和再现移动性的异构特征。我们评估我们的模型并将其性能与合成和真实数据测试用例上的 WHERE 算法进行比较。使用 3W 模型将使用 Earth Mover's Distance 测量的人口分布再现精度提高了 35%。3W 模型再现的流动距离和空间分布与输入数据高度匹配。我们还通过比较合成数据集和输入数据集来评估隐私保护水平。