SRC: Discovering human activity community in a city

Abstract

This study investigates human activity community in a city by conceptualizing it as a network embedding problem. In order to learn the latent representations of activity-travel patterns from individual daily trajectories, network embedding learns a vector space representation for each type of activity place as a node connected by movement links to preserve the structure of individual activities. The proposed approach is applied to mobile positioning data at the individual level obtained for a weekday from volunteers at Guangzhou City. Assessments are conducted to validate individual decision making for several types of activities by a field survey. This study contributes to a general framework for discovering individual activity-travel patterns from human movement trajectories.