Abstract: To address the challenges of managing and efficiently localizing large-scale, multi-category urban spatiotemporal data, an interactive visu-alization method based on point aggregation was proposed. First, efficient aggregation algorithms based on geographic location and geo-graphic hierarchy were introduced to address the needs of government personnel for effective visualization management in different scenarios. Second, a conditional parsing algorithm was proposed based on efficient point aggregation to enable real-time parsing and conversion of spatiotemporal conditions and category visibility, improving data localization efficiency. The method used nearly 270,000 urban entities in Beijing to conduct geographic hierarchical parsing algorithms and aggregation interaction experiments. The experiments demonstrated that the average performance of our two aggregation methods exceeds improvements of 70% and 60% over the K-means in various scenarios, confirming the system's efficiency and stability in data processing and aggregation services. This aggregation application has been successfully implemented in the Yizhuang urban governance project, supporting large-scale urban data aggregation for millions of entities and highlighting the high feasibility of the system.

Key words: Point Aggregation, Geographical level analysis, Human-computer Interaction, Visualization Management System, Spatiotemporal Data

摘要： 摘 要: 针对大规模多类别城市时空数据可视化管理难、定位效率慢的问题，提出基于点聚合的多类别城市时空大数据交互式可视化方法。首先，分别提出基于地理位置与基于地理层级的高效聚合算法满足政务人员在不同场景下的高效可视化管理；其次，在高效点聚合功能基础上，提出条件解析算法实现对时空条件、类目显隐的实时解析转换，提高数据定位效率；最后，采用北京市近27万城市实体数据进行地理层级解析算法与聚合交互实验。实验中不同场景下，所提出的2个聚合方法的平均性能相对于K-means方法均分别提高超过70%和60%，充分说明了系统存储数据时数据处理与聚合服务的高效性与稳定性；同时该聚合应用服务目前已于北京市亦庄城市治理项目示范应用，支撑百万级城市数据聚合服务，体现该应用设计具有较高的可用性。

关键词: 点聚合, 地理层级解析, 人机交互, 可视化管理系统, 时空数据