基于网格化管理事件大数据的上海市气象与城市运行体征关联规则挖掘  

Mining of association rules between meteorological and urban operation signs based on grid management events big data in Shanghai City

作　　者：杨辰 王强 金诚 李海宏 任洪润 Yang Chen;Wang Qiang;Jin Cheng;Li Haihong;Ren Hongrun(Shanghai Meteorological Service,Shanghai 200030,China;Shanghai Center for Meteorological Disaster Prevention Technology,Shanghai 200030,China;School of Computer Science,Fudan University,Shanghai 200433,China;Shanghai Key Laboratory of Data Science,Shanghai 200433,China)

机构地区：[1]上海市气象局,上海200030 [2]上海市气象灾害防御技术中心,上海200030 [3]复旦大学计算机科学技术学院,上海200433 [4]上海市数据科学重点实验室,上海200433

出　　处：《地理科学》2024年第5期874-882,共9页Scientia Geographica Sinica

基　　金：上海市2020年度“科技创新行动计划”社会发展科技攻关定向项目(20DZ1200403)资助。

摘　　要：基于时空特征分析、共现词项特征分析、相关性分析及频繁模式挖掘方法对城市运行管理大数据进行分析研究,得到触发网格化管理事件发生的典型气象条件,并构建涵盖气象条件的典型网格化管理事件知识图谱。结果表明,网格化管理件发生时间与工作时间高度吻合,发生区域也与城市人员密集区相重合,类别上存在“头部集中、长尾分布”的现象,网格化管理事件分词上可以形成较为清晰的聚类结构,形成以市民活动为主体的共现词项关系网络。结合气象资料分析,市政设施、环卫等小类与气温相关性较为明显,风易损结构受风力影响较大,并且在降水、低温、高温和大风等特定天气情况下基坑、纠纷类、高空抛物和河道绿化等事件将呈现高发趋势。此外,通过采用知识图谱技术归纳和表达气象与城市运行之间的关联,从而有利于城市运行管理人员在特定天气条件的提前应对和处置。Refinement governance is the future governance direction of the city,and it is also an important challenge for Shanghai to build an outstanding global city.Most of the existing research is based on the innovative mode and management mechanism of urban grid management,but the analysis and mining of grid management event data are still insufficient,and there is a lack of regular analysis of meteorological conditions on the occurrence of events.From the meteorological perspective,this paper uses spatiotemporal feature analysis and natural language processing methods to analyze the features of grid management event data,and uses correlation analysis and frequent pattern mining algorithms to obtain the association rules between meteorological conditions and urban management events.On this basis,the typical meteorological conditions that trigger grid management events are obtained,and the typical event knowledge graph covering meteorological conditions is constructed.The results show that the events are highly correlated with the characteristics of residents’activities,the occurrence time of events is highly consistent with the working time,and the occurrence area also coincides with the densely populated areas of the city.There is a phenomenon of“concentrated head and long tail distribution”in the category,and a clear clustering structure can be formed in the event word segmentation,forming a co-occurrence term relation network with citizen activities as the main body.Analysis with meteorological data,municipal facilities and sanitation categories have obvious correlations with air temperature,wind-vulnerable structures are greatly affected by wind,and some illegal behaviors are also highly correlated with meteorological conditions.In addition,under specific weather conditions,some events will show an obvious tendency to occur easily.For example,events such as foundation pits,disputes,high-altitude parabolas,and river greening occur under specific weather conditions such as precipitation,low temperature,high tem

关 键 词：网格化管理 气象 特征挖掘 FP-GROWTH 知识图谱 

分 类 号：P429[天文地球—大气科学及气象学]