A spatiotemporal knowledge graph-based method for identifying individual activity locations from mobile phone data
Spatiotemporal knowledge graph framework for activity-location identification. Source: paper.Abstract
This study proposes a spatiotemporal knowledge graph method for identifying individual activity locations from mobile phone data. Spatial adjacency and temporal co-occurrence are modeled as separate graphs, fused into a weighted spatiotemporal graph, and partitioned with Louvain community detection. A Shanghai case study shows tighter spatial boundaries and more stable temporal identification than conventional threshold-based and spatiotemporal clustering methods.
Publication
Journal of Transport Geography, 124, 104157
研究概览
本研究针对手机信令数据时间稀疏、空间定位不确定以及传统聚类方法依赖人工阈值的问题，将个体轨迹组织为时空知识图谱。方法同时推断停留点的空间邻接关系和时间相似关系，再识别紧密关联的活动地点群。
核心方法与发现
- 用三元组表示个体、停留点、时间片及其关系，形成可解释的轨迹语义结构。
- 基于 Queen 邻接构建空间图，基于时间片余弦相似度构建时间图，并将两者融合。
- 使用 Louvain 社区发现自动识别活动地点，在上海案例中改善空间边界精度和日间稳定性。
资料来源
Jian Li
Professor
Tian Gan
Ph.D. Student
