基于概率主题模型的网络定位结果优化  

Optimization of Network Geolocation Results Based on Probability Theme Model

作　　者：张龙[1] 周杨[1] 吕亮[1] 赵帆[1] 梁静 ZHANG Long;ZHOU Yang;LYU Liang;ZHAO Fan;LIANG Jing(Information Engineering University,Zhengzhou 450001,China;Henan College of Surveying and Mapping,Zhengzhou 450015,China)

机构地区：[1]信息工程大学,河南郑州450001 [2]河南职业测绘学院,河南郑州450015

出　　处：《测绘与空间地理信息》2018年第8期31-35,共5页Geomatics & Spatial Information Technology

基　　金：国家重点研发计划(2016YFB0801301;2016YFB0801303)资助

摘　　要：现有基于网络测量的IP定位方法往往难以给出带有准确经纬度信息的可靠定位结果,通常仅能确定目标可能所处的大致区域(本文称为缓冲区),定位精度难以满足实际应用需求。针对该问题,本文提出了一种基于概率主题模型的网络定位结果优化方法。该方法首先从爬取的与网络实体相关的文本信息内容中,利用概率主题模型LDA的弱监督主题分类思想,提取与地物类型相关的从属主题;其次,根据提取的从属主题,确定实体可能所属地物类型;最后,将网络定位结果缓冲区与地理图层进行叠加,在叠加范围内检索所属类型的地物,确定实体的地理位置,从而完成文本主题与地物类型的匹配,实现对原有定位结果的优化。对仿真数据的测试结果表明,该方法能够在原有网络定位结果的基础上进一步缩小网络实体资源所处空间分布范围,证明了该方法的有效性。Existing IP location method based on network measurement is often difficult to give reliable geolocation result with accurate longitude and latitude information. It usually can only determine the approximate area（ it called as buffer area in this paper）. And the precision is difficult to meet the practical application requirements. Aiming to resolve this problem,this paper proposes an optimization method of network geolocation results based on probability topic model. First,from the content of textual information related to network entities,we use the weakly supervised thematic classification idea of the probabilistic thematic model-LDA to extract subordinate topics related to landmark types. What’s more,according to the subtopic extracted,the type of landmarks that may belong to the entity is identified. In the end,we computes the superposition of network geolocation buffer and geographic layer,where the landmarks can be indexed to further determine the geographic location of the entity. Thus the textual topics and the landmarks categories are matched and correlated,and the location of the entity is determined to perform the optimization of the original geolocation results. Test result shows that this method can narrow space distribution of network entity resources comparing with the results of the original network geolocation,and the effectiveness of the proposed method is proved.

关 键 词：IP定位 LDA GIBBS采样 缓冲区分析 

分 类 号：P208[天文地球—地图制图学与地理信息工程]