Data-driven generation of spatio-temporal routines in human mobility.

Area of Science:

• Computational Social Science
• Mobility Modeling
• Human Behavior Simulation

Background:

• Realistic human mobility simulation is crucial for applications like network protocols and urban planning.
• Existing generative models struggle to capture both routine adherence and routine-breaking behavior in human movement.

Purpose of the Study:

• To present Ditras (DIary-based TRAjectory Simulator), a new framework for simulating spatio-temporal human mobility patterns.
• To develop a data-driven approach that accounts for routine and deviations in human mobility.

Main Methods:

• Ditras generates mobility by first creating a 'mobility diary' and then translating it into a trajectory.
• A data-driven algorithm constructs the diary generator, capturing routine adherence and deviations.
• A trajectory generator based on preferential exploration and return mechanisms is proposed.

Main Results:

• The proposed Ditras algorithm, using its data-driven diary and preferential exploration/return generators, demonstrated superior accuracy in reproducing statistical properties of real human trajectories.
• Comparative analysis showed Ditras outperforms other generative algorithms and synthetic data generators.

Conclusions:

• Ditras provides a significant advancement in simulating realistic human mobility patterns.
• The framework enhances the understanding of the origins of spatio-temporal human mobility patterns by accurately modeling routine and deviations.