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.