近年，城市地理学和社会地理学领域对流空间分异的关注日益增加。流空间分异相较基于地点和基于人的分异，主要的区别在于过程性。然而，基于流空间的社会分异研究仍处于理论和方法的探索阶段，亟待研究人员探索。本研究基于手机信令数据，计算每条通勤流上各收入群体间的接触指数；统计每个空间网格相关属性，基于深度嵌入聚类算法对社会空间划分。观察各类收入群体通勤流的暴露状况，探索社会空间类型及分布位置。研究表明，共有的通勤大通道是通勤过程中群体间维持接触的媒介；不同收入群体更大概率与收入接近的群体发生接触，从而减小整体的分异；通勤移动使广州形成了“一核多中心”的圈层结构。

In recent years, there has been growing interest in the flow space differentiation within urban and social geography. Flow space differentiation, distinct from location-based and person-based differentiation, is characterized by its process-oriented nature. However, research on flow space differentiation is still in its early stages regarding both theory and methodology, necessitating further exploration by researchers. This study, leveraging mobile signaling data, calculates the exposure index between different income groups on each commuting flow. The research conducts a statistical analysis of attributes associated with each spatial grid and employs a deep embedding clustering algorithm to delineate social spaces. The primary objective of the study is to observe the exposure levels of various income groups during commuting flows and explore the types and distribution of social spaces. The study indicates that common commuting corridors serve as mediums for maintaining contact between different groups during commuting. Different income groups are more likely to come into contact with groups closer to their income levels, thereby reducing overall differentiation. Commuting movements have led to Guangzhou forming a concentric "one core, multiple centers" structure.