Incorporating activity content knowledge for identifying human mobility pattern

Identifying meaningful patterns of human mobile from accumulating trajectory is essential for understanding human behaviors. However, previous works identify human mobility patterns based on the spatial co-occurrences of trajectories, which ignores the knowledge of activity content, remaining challenges in effectively identifying and understanding mobility patterns. To bridge this gap, this study introduces the knowledge of activity context for discovering human mobility patterns from trajectories. The proposed model, knowledge-aware human activity pattern model (KHAP), first embeddings the knowledge of activity content in distributed continuous vector space by taking POI as an agent, then discovers representative and interpretable activity patterns from human trajectory sets by an unsupervised approach. Finally, several evaluation metrics are derived to examine human activity patterns, including pattern coherence, pattern similarity, and manual scoring. A real-world case study is conducted to examine the performance of the proposed model, and the experimental results show the proposed model improves interpretability and helps researchers understand activity patterns. This study not only offers a novel solution for identifying human mobility patterns, but also provides a method reference for fusing content semantics of human activities for trajectory analysis and mining.