众源地理信息的可信度计算模型与算法研究

Data quality of crowd-sourcing geographic information varies greatly because of its openness and unprofessional volunteers. Although there is a sophisticated mechanism for evaluating the quality of traditional geographic information, it cannot apply to the quality assessment of crowd-sourcing geographic information in such an open environment. It is really an important problem, which arrests the attention of the domain of geographic information science, to explore a suite of effective quality assessment to adapt to crowd-sourcing geographic information. Fortunately, the trustworthiness-based quality assessment mechanism is a feasible alternative. This project aims at this international academic frontiers and will research models and algorithms of trustworthiness computation for crowd-sourcing geographic information. First of all, after researching the complementariness of trustworthiness measurement between professional geographic information and the new environment, the influencing mechanism of trustworthiness of crowd-sourcing geographic object will be analyzed. Secondly, by means of exploring and investigating principle and computing method for implicit assessment of contributors by amendment and change information of geographic objects, a dynamic computation model of synthetic volunteer reputation based on implicit assessments among volunteers for crowd-sourcing geographic information will be proposed. At last, the conflict degree between geographic objects and their surrounding objects will be concentrated on. And a crowd-sourcing geographic information trustworthiness model integrating with volunteer reputation, quality of crowd-sourcing geographic information and conflicts of geographic objects will be come up. This study will deepen theory foundation of trustworthiness computation of crowd-sourcing geographic information and provide innovative ideas and theories for the quality control and screening of crowd-sourcing geographic information.

众源地理信息的开放性与贡献者的非专业性导致众源地理数据质量参差不齐。尽管传统地理数据质量评价有一套较完善的体系，但难以适用于众源开放环境下的地理数据质量评价。探索一套适应众源数据的有效质量评价方法，是地理信息领域十分关注的重要问题。本项目拟针对这一国际学术前沿问题，开展基于众源地理信息可信度的质量评价机制研究：(1) 研究众源地理信息与专业地理信息在可信度计算模式上的互补性，系统地分析众源地理信息可信度的影响机理；(2) 应用地理目标修正与变更数据，探索研究贡献者多类隐式评价的原理和量化方法，构建基于隐式评价的众源信息志愿者综合信誉动态计算模型；(3) 研究地理目标与周边地物的相容度量化方法，提出融合基于志愿者信誉、目标相容度和目标自身质量等多类信任关系的众源地理信息可信度计算的综合模型。预期研究成果将深化众源地理信息可信度计算的理论基础，为众源信息的质量控制和筛选提供新的思路和理论依据。

众源地理信息志愿者的非专业性和贡献平台的开放性是导致众源地理数据质量问题的主要因素，针对传统地理数据质量评价方法难以适应众源地理信息新环境，本项目开展基于众源地理信息可信度的质量评价机制研究，主要在以下六个方面取得了重要研究成果：（1）研究了众源地理信息志愿者贡献特征等影响地理信息可信度的因素，提出了基于改进WPCA的特征降维和用户分类方法，为贡献者信誉评价奠定基础；（2）研究了贡献者修改众源地理目标过程中要素类标签的变迁情况，提出利用图结构对其建模的方法，并采用基于Louvain社区发现方法挖掘并分析要素类标签的变迁模式；（3）提出了基于xDeepFM的众源地理信息要素类标签推荐方法和基于改进马尔可夫链的非要素类标签推荐方法，为众源地理信息质量和可信度提升提供了新思路；（4）提出了基于众源轨迹数据的道路组成要素识别方法，通过深度学习识别道路、交叉路口和红绿灯等，为多源数据融合和交叉验证提供了良好的基础；（5）设计了众源地理信息本体，结合版本间的语义相似度更加全面地隐式评价用户贡献的地理目标质量，并提出了基于本体语义相似度的贡献者信誉动态评价方法；（6）提出了结合用户信誉用户信誉、目标本身质量和目标相容度的众源地理信息可信度评价方法，为众源地理信息的质量控制和筛选提供新的思路和理论依据。基于相关研究成果发表论文27篇，出版专著1部，其中被SSCI/SCI索引6篇、EI收录12篇，申请国家发明专利3项，其中1项授权、2项实审，申请软件著作权登记4项，培养硕士研究生10名。