顾及地理信息云服务领域知识的空间分析任务日志识别方法  

A Spatial Analysis Task Log Recognition Method Considering Domain Knowledge of Geographic Information Cloud Service

作　　者：李江 刘朝辉 宋旭颖 李锐[1] 吴华意[1] LI Jiang;LIU Zhaohui;SONG Xuying;LI Rui;WU Huayi(State Key Laboratory of Information Engineering in Surveying Mapping and Remote Sensing,Wuhan University,Wuhan 430079,China;Information Center of Department of Natural Resources of Hubei Province,Wuhan 430071,China)

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]湖北省自然资源厅信息中心,湖北武汉430071

出　　处：《测绘地理信息》2023年第1期86-92,共7页Journal of Geomatics

基　　金：国家自然科学基金(U20A2091,41771426);湖北省科技攻关计划(ZRZY2021KJ13)。

摘　　要：现代地理信息云服务平台在处理高强度空间分析事件的同时,通常依据到达时序记录来自不同用户、围绕不同分析目的的事件日志。这种混杂的日志记录方式模糊了用户的分析意图,破坏了围绕同一分析目的的计算行为的时间关联性,削弱了服务日志对从用户需求角度理解用户行为和优化服务的重要作用。提出了一种顾及地理信息云服务领域知识的空间分析任务日志识别方法,对围绕同一分析目的的服务日志进行自动化聚合。首先,基于历史任务数据集对地理信息云服务中的日任务量时序变化规律和图层关联性等领域知识进行建模;然后,借助层次编码和超参数设定的方式建立基于领域知识的聚类经验约束;最后,通过k-means聚类及后处理得到空间分析任务日志的识别结果。基于地理信息云服务平台产生的大量空间分析日志进行实验,结果显示,利用所提方法对空间分析任务日志进行识别,F1值可达到0.895,相比无领域知识支持的基线方法,其整体精度提升了8.7%以上,可有效提高空间分析任务日志识别精度。While dealing with high-intensity spatial analysis events,the modern geographic information cloud service platforms usually record data from different users with different analysis purposes based on the arrival sequences.This mixed log recording method obscures the users’analysis intention,destroys the time correlation of computing behaviors with the same analysis purpose,and weakens the importance of platform log in understanding users’behaviors and optimizing service from the perspective of users’needs.Therefore,we propose a spatial analysis task log recognition method considering domain knowledge of geographic information cloud service to automatically aggregate the service logs with the same purpose.First,based on the historical task data set,we model the time series change rule of daily tasks and the layer correlation in the geographic information cloud service.Then,we establish the clustering experience constraint with domain knowledge by means of hierarchical coding and hyper-parameter setting.Finally,the recognition results of spatial analysis task logs are obtained by k-means clustering method and postprocessing.Experiments are carried out based on a large number of spatial analysis task logs generated by the geographic information cloud service platform.The results show that the F1 value can reach 0.895 when using the proposed method to recognize the spatial analysis task logs.Compared with the baseline method without domain knowledge intervention,its overall accuracy is improved by more than 8.7%,which can effectively improve the recognition accuracy of spatial analysis task logs.

关 键 词：领域知识 任务识别 日志聚合 地理信息云服务 

分 类 号：P208[天文地球—地图制图学与地理信息工程]