Quantitative Assessment of City’s Smart Growth Level in China Using Fuzzy Comprehensive Evaluation

. In this paper we develop an indicator system to evaluate the Smart Growth Level based on 3E’s principles and calculate the Smart Growth Level of 70 mid-scale and large cities in China using fuzzy comprehensive evaluation model. We also divide these cities into four grades: outstanding, good, normal and terrible. The results are significant for governments to design a reasonable development plan.


Introduction
Urban Sprawl is low-density, non-continuous, automobile-dependent residential and non-residential development model, which needs to cost a remarkable amount of land and other natural resources. In china, this unsustainable development model has affected many cities recently, which brought several problems with environment, economy as well as society [1]. Smart growth is an urban planning theory that originated in 1990's as a means to curb continued urban sprawl and reduce the loss of farmland surrounding urban centres, which focuses on building cities that embrace the E's of sustainability-Economically prosperous, socially Equitable, and Environmentally Sustainable [2]. Therefore, the problem how to use the checking data to evaluate the Chinese cities' Smart Growth Level (SGL) has been the public concern.
Badill et al illustrated the work in the context of Sii-Mobility smart city project, which is to define a smart city architecture. A wide range of processes and data is presented [3]. Anna Luusua et al introduced a new way of designing called nonanthropocentric design to decenter the human as the focus of design while the smart city agenda is critique for its focus on technology and business led solutions [4]. Chang Saisai et all proposed a fuzzy comprehensive evaluation model based on Analytic Hierarchy Process(AHP) to judge whether a city's growth plan is reasonable [5].
In this paper, we studied the SGL of 70 large and mid-scale cities in China to find the one that the City Sprawl is the most serious, which can give some constructive suggestions when the government makes a new development plan.

Data Collection
Collecting sufficient data is the basis of developing a complete index system. We searched the China Economic and Social Development Statistics Database [6] and found 22 indicators of 70 mid-scale and large cities firstly based on 3E's principles.

Data filling
It is important to ensure the continuity and authenticity of research data. However, some data is missing because not all data is provided in the website. To address this problem, two methods are proposed to complete the data, which are as follow,

Average value
If we can found the data of an indicator for most of cities except few cities, the average value of other cities can be taken as the missing. For example, the number of public libraries in Dali is missing so we adopt the average value of other 69 cities to replace it. As these figures are really few, it affects little to the the evaluation result.

GM(1,1) model
As the frequently used grey model, GM(1,1) is particularly available to such random unconfirmed issue. If the data before can be obtained, GM(1,1) Model is adopted to predict the figures we need. For example, the data for Rural Subsistence Expenditure, Total Import and Export, Urban Social Security Engagement Rate, Expenditure on Education Expenditure, Number of Health Institutions in Hefei in 2015 is missing while the x k x i k n = = =  and the sequence mean generated of consecutive is called a grey differential equation, and is also called the model GM(1,1). The equation The result is shown in Figure.      The results show that using the GM(1,1) model to forecast the missing data involves a smaller sample and less information. It's easier to build the model and the prediction precision is better.

Quantitative assessment of SGL
We apply the Minimum Relative Deviation Distance method in comprehensive evaluation model as this way does not require data pre-processing.

Set up the relative deviation fuzzy matrix
The ideal alternative is

Evaluation Model Based on Variation coefficient method
In this paper, variation coefficient method is adopted to calculate the weight vector in this section. The coefficient of variation of each index is   The larger SGL is, the development model of given city meets the requirements of Smart Growth better.

The Standard of Smart Growth
According to the SGL evaluation model aforementioned, we can get the standard of smart growth using K-means clustering method. Then the mean of the indicator centres is used as the standard boundary.
In this way, 70 cities are divided into for kinds: outstanding, good, normal and terrible. Terrible means that the development model has limited and threated the further development of given city. Normal means that the city has not realized smart growth in this area. And good means that the city has some success of smart growth in this area. Outstanding is used to describe when city has great success of smart growth. The standard of SGL and the classification of 70 mid-scale and large cities in China is shown in Table.3. and

Conclusion
There exits a great difference in the model and level of cities' development in China. Specifically speaking, some cities like Beijing and Shanghai have achieved great success of smart growth while some of others like Yueyang and Lanzhou are suffering from a bottleneck period in pursuing development. Positive measures and smart growth plan should be adopted to help these cities to achieve smart growth.