A knowledge graph-based approach for recognizing older adults’ behavior in urban parks

The behavior of the older adults in urban parks could greatly reflect the behavioral traits of the older adults and their interactions with the environment. Therefore, effective and efficient identification of older adults’ behaviors in urban parks is important for exploring their outdoor behavioral patterns and assessing the quality of age-friendliness urban parks. In this research, a knowledge graph (KG)-based method for recognizing older adults’ behavior was innovatively proposed. The proposed approach was based on the KG of older adults’ behavior (OBKG) established from the image datasets collected via Baidu street view and social media platforms. Additionally, it was also coupled with the link prediction to complete the inference prediction of the older adults’ behavior based on the entities and relationships within the images, which can help improve the accuracy of the recognition of older adults’ behavior. The results of the research showed that based on the YOLOv5 model, our proposed method could improve the behavior recognition accuracy by 13.9%, and the improvement of group behavior recognition accuracy could reach nearly 53.9%. Lastly, the potential applications, implications of this proposed approach were also discussed, alongside the limitations and the future direction of our research.