Research On VGI Data Quality Assessment Based On Neural Network And Fuzzy Inference

With the constant development of data science and computer technology,the traditional mode of data dissemination has gone through profound changes driven by the internet big data.As a new type of geographic information in the era of Web 2.0,Volunteered Geographic Information(VGI)enables users to become contributors of geographic data through a crowdsourcing way.Under such circumstances,the complex sources and uncertain propagation paths of VGI lead great impacts on its data quality.Therefore,applying reasonable mathematical methods to characterize and evaluate the quality of data can provide users with important theoretical basis and useful references.In this dissertation,the characteristics of VGI data are analyzed,and a general theoretical framework to evaluate the quality of VGI data is also proposed.Based on this analysis and proposal,a probabilistic method and neural network method are used to calculate the positional accuracy and semantic trust of VGI data.Finally,we built an indicator of credibility based on fuzzy inference theory to characterize the quality of data and apply it into the real data in Kowloon,Hong Kong.The main works of this dissertation are listed as follows:(1)This dissertation studies the characteristics of VGI data.To deal with the positional information and semantic information of VGI data,we propos a framework of general quality assessment that can combine both accurate and inaccurate information into a single measure by analyzing its data connotation;(2)We establish a probabilistic assessment of VGI positional accuracy according to the characteristics of VGI and drawing on the error measurement method of GIS;(3)We improve and modify the trust model of existing data theoretically,and then use the neural network method to calculate the semantic trust of the data according to the characteristics of VGI data;(4)We combine the positional accuracy with semantic trust of VGI data in order to construct a quantitative quality indicator called “Credibility” based on a fuzzy inference system.Moreover,methods of constructing the corresponding membership functions and fuzzy rules are proposed.Finally,an integrated model was built to assess the quality of data in practical.