A Cognitive Approach to Modeling Vague Geographical Features: A Case Study of Zhongguancun

LIU Yu,YUAN Yi-hong,ZHANG Yi

DOI: https://doi.org/10.11834/jrs.20080249

2008-01-01

Journal of Remote Sensing

Abstract:In practice,vagueness is a common phenomenon of geographical features.The vagueness of a feature often comes from human conceptualization of the real world.Modeling of vague features will undoubtedly contribute to more precisely handling spatial knowledge.In recent studies,a number of theoretical methods have been employed to model vague features,where fuzzy set theory is in common use.Following that theory,the degree that an element belongs to a fuzzy set can be expressed by a number between 0 and 1.We can thus establish a corresponding membership function(MF) for a fuzzy set.Recently,much literature focuss on vague features and proposes approaches to establish corresponding MFs.They include approaches based on cognitive experiment,remotely sensed data and GIR(Geographical Information Retrieval).Due to the subjectivity of vagueness,spatial cognitive experiments provide a direct way to represent the vagueness of individuals' conceptualization of corresponding features.However,previous methods based on cognition cost highly and are somewhat hard to control the result,since subjects in such experiments are asked to delineate boundaries of vague areal objects.Landmarks play an important role in individuals' development of spatial cognition.It is thus relatively easy for individuals to perceive a landmark and decide whether it is within a given region or not.In this research,we took Zhongguancun in Beijing city as an example,since it is complex with different meanings,such as educational,political,and historical meanings.A questionnaire is designed to collect membership degrees of 30 landmarks which are in the region of Zhongguancun.These 30 landmarks,which are selected from the maximum potential region corresponding to Zhongguancun,can be abstracted to point features.They belong to different types,such as office building,hotel,school,recreation place,and natural feature.For each landmark,the subjects are asked whether it is within Zhongguancun,for which three optional answers are provided: YES,NO,and NOT SURE.By collecting all answer sheets,we can compute a score of each landmark.Such a value can be viewed as the membership degree that the corresponding position belongs to the concept of "Zhongguancun".However,since Zhongguancun is a two-dimensional vague object,it should be represented by a membership function(MF) like z=f(x,y).We thus need to find out an appropriate interpolation method to obtain the MF.In this research,support vector regression(SVR) is adopted to compute the MF.Compared with conventional interpolation methods,such as IDW(Inverse Distance Weighted) method,the proposed approach is easy to implement and the results are convenient to be managed.Additionally,SVR provides a mechanism to obtain a trade-off between goodness of fit and generality by adjusting some parameters,such as γ for radial basis kernel functions.Based on the result of membership function,we also investigate spatial distribution properties of Zhongguancun and find out some interesting points.Since Zhongguancun has been viewed as "the silicon valley in China",it is closely related with such concepts as high-tech industry,university,and so on.Consequently,the landmarks associated with these concepts always have higher membership degrees.On the contrary,lower scores are assigned to some recreation places and natural features.In summary,from a point of view of behavior,the current concept of Zhongguancun is far beyond the scope as an administrative unit,both spatially and functionally,since many factors influence its internal representation of individuals.