Resources Policy 49 (2016) 394–404

Contents lists available at ScienceDirect

Resources Policy

http://d0301-42

n CorrE-m

journal homepage: www.elsevier.com/locate/resourpol

Analyzing sustainability of Chinese mining cities using an associationrule mining approach

Lijun Zeng a,n, Bingcheng Wang a, Liu Fan a, Jianguo Wub

a College of Economics and Management, Shandong University of Science and Technology, 579 Qianwangang Road, Qingdao 266590, PR Chinab School of Life Sciences and School of Sustainability, Arizona State University, Tempe, AZ 85287, USA

a r t i c l e i n f o

Article history:Received 4 April 2016Received in revised form26 July 2016Accepted 26 July 2016

Keywords:SustainabilityMining citiesAssociation rulesData miningChina

x.doi.org/10.1016/j.resourpol.2016.07.01307/& 2016 Elsevier Ltd. All rights reserved.

esponding author.ail address: zenglijun518@126.com (L. Zeng).

a b s t r a c t

Mining cities are of vital importance to the global development, but they are facing great challenges toachieve sustainability goals. Understanding the characteristics and patterns of the environmental, eco-nomic, and social conditions of mining cities is critical to promoting their sustainability. After developingan Indicator System for Mining City Sustainability, we collected data from 110 prefecture-level cities andanalyzed environmental, economic, and social characteristics of these mining cities using a data-miningmethod – Association rule mining. Our analysis revealed some novel, implicit, and previously unknowncharacteristics and patterns of mining city sustainability in China. We found that education investment,economic development, and some aspects of society were substantially unbalanced in most Chinesemining cities. Most coal-mining cites had a larger proportion of mining population, and usually facedsevere challenges in reducing industrial dust emission. The unsustainable characteristics of Chinesemining cities exhibited distinctive regional patterns and should be considered explicitly in policy makingto promote the sustainability of these urban areas.

& 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Mining cities have played an important role in socioeconomicdevelopment around the world. In China, mining cities provideprimary raw materials and make great global contributions. Since1949, Chinese mining cities have provided 52.9 billion tons of rawcoal, 5.5 billion tons of crude oil, and 5.8 billion tons of iron ore forthe whole country. However, these mining cities face great chal-lenges in socioeconomic development and ecological environmentprotection because of their slow undertakings of social develop-ment, imbalanced economic structure, and seriously damagedecological environment. To promote the sustainable development ofmining cities, the first and most important step is to identify clearlythe patterns and characteristics of mining city sustainability.

Several research efforts have been made in mining sustain-ability following the rising importance of sustainability. On onehand, mining activities have much in common with industryactivities worldwide. Consequently, several researchers have con-tributed to the evaluation of mining sustainability from a miningindustry perspective, mainly including evaluation indicator sys-tems and evaluation methods. Among the burgeoning evaluation

indicator systems, the Global Reporting Initiative (GRI) frameworkand its mining and metals sector supplement (MMSS) is arguablythe most widely adopted evaluation system in the mining industry(GRI, 2000, 2012). Based on the second generation of GRI, AdisaAzapagic developed an evaluation framework for the mining in-dustry, which has been cited widely (Azapagic, 2004). In addition,many industrial associations and organizations, such as the MiningAssociation of Canada (MAC), Mining, Minerals, and SustainableDevelopment project, and several researchers including Yu (2005,2008) and Giurco (2012) also proposed evaluation indicator sys-tems for mining industry sustainability (Yu et al., 2005, 2008;Giurco and Cooper, 2012). Moreover, Worrall discussed an eva-luation system for legacy mine land while Solomon studied thesocial dimension of mining sustainability (Solomon et al., 2008;Worrall et al., 2009). With respect to evaluation method, Yu (2005,2008) applied the fuzzy integrated judgment model and factoranalysis (Yu et al., 2005, 2008), Kommadath (2012) discussed afuzzy logic-based approach, and Valdivia (2011) utilized a life-cy-cle inventory method (Kommadath et al., 2012; Valdivia andUgaya, 2011).

On the other hand, mining sustainability is closely related tothe condition of the special geographical space where mining ac-tivities are carried out. Different geographical spaces have differ-ent ecological environmental, economic, and societal conditions;therefore, the patterns, characteristics, strategies, and policies for

Econoomy

Ec−−

Economy−Environme

Environmen

conomy−Soc−Environme

−E

Economy−Society

yent

So

nt

ietynt

SocietyEnvironmen

ociety

nt

Fig. 1. Improved Triple Bottom Line Model.

L. Zeng et al. / Resources Policy 49 (2016) 394–404 395

mining sustainability may vary greatly in different locations. As aresult, a number of special researches and case studies wereconducted on different mining locations using different scales,such as mining site, mining city, country, and region. Bebbingtonet al. (2009) studied the institutional challenges for mining andsustainability in Peru. They found that mining expansion hasplaced pressures on water resources, livelihood assets, and socialrelationships, and these pressures are because of the institutionalconditions that separate the governance of mineral expansion,water resources, and local development, and of asymmetries ofpower that prioritize large-scale investment over livelihood andenvironment (Bebbington and Bury, 2009). Mudd (2007) ex-amined the long-term trends of almost all sectors of the Australianmining industry by illustrating the principal issues concerningmining and sustainability: increasing production, declining oregrades, increased open-cut mining and associated waste rock oroverburden and remaining economic resources. The author con-cluded that the sustainability of the mining industry in Australiacontinues to hang in the balance (Mudd, 2007). Hajkowicz et al.(2011) examined the relationship between mining and socio-economic wellbeing in Australia and found that mining activityhas a positive impact on incomes, housing affordability, commu-nication access, education and employment across regional andremote Australia. Sorensen (2012) studied the sustainable devel-opment of mining companies in South Africa and pointed out theirinconsistent understanding of the idea of sustainability.

In addition, mining activities are usually organized in the formof projects, and a number of studies were from the perspective ofmining projects and sustainability. McLellan et al. (2009, 2012,2013) wrote a series of papers on integrating sustainability intodesign and management of mining projects (McLellan et al., 2009;Corder et al., 2012; Tuazon et al., 2012; McLellan and Corder, 2013).Phillips (2012) applied a mathematical model of sustainability tothe environmental impact assessment of a proposed bauxite-mining project in Indian and indicated the potential of this modelin problem resolutions of sustainability of local mining projects.

Considering the complexity and systematic nature of miningsustainability, all these research perspectives are necessary andsignificant. Research from a sustainable mining industry perspec-tive can promote information exchange and comparisons betweenmining corporations to contribute to mining sustainability. Re-search from a sustainable mining projects perspective can helpimplement sustainable mining activities according to particularcircumstances, while research from a mining region perspectivecan comprehensively and profoundly identify the sustainability ofthe mining region and the relationships between mining and localsustainability. Considering that mining activities have funda-mental influences on the local environment, economy and society,the mining region perspective is ideal and greater attention needsto be paid to this perspective. Regarding the scale of the mininglocation, a national scale may be too large to understand the localeffects of mining on sustainability, whereas a site scale may be toosmall to include all the important effects of mining. Thus, weconsider a mining city to be a key scale for studying mining sus-tainability. Overall, the study of mining city sustainability is still inits infancy and needs further research.

The main purpose of this study was to investigate the environ-mental, economic and social characteristics of mining cities in Chinausing association rules mining, which is a novel method in this field.Following the Tipple Bottom Line model of sustainability, we de-veloped an Indicator System for Mining City Sustainability (ISMCS),collected data from 110 prefecture-level mining cities, and appliedAssociation rule mining to the collected dataset. The mined asso-ciation rules and frequent patterns were analyzed to identify thepatterns and characteristics of mining city sustainability in China, soas to provide a scientific basis for policy making.

2. Methods

2.1. Indicator system for mining city sustainability

We propose the ISMCS based on the triple bottom line model(Fig. 1). As we can see from the triple bottom line model, apartfrom the three fundamental dimensions, environment, society andeconomy, there are three intersections of two dimensions and oneintersection of three dimensions. Here the four intersections aredifferentiated from the three fundamental dimensions and treatedas separate dimensions. That is, in addition to the three funda-mental dimensions (environment, economy, and society), thereare three integrated two-dimensional domains (economy-society,economy-environment, society-environment), and one integratedthree-dimensional domain (economy-society-environment). Thisstructure will be helpful not only to represent and understand thesustainability status of mining cities in China, but also to analyzeand interpret the mined association rules. Furthermore, for miningcity in China, four type indicators is vital to both the description ofmining city status and discovery of characteristics and patterns ofmining city sustainability. Among the four indicators, Province andRegion describes each mining city's geographical location, Stagerepresents each mining city's development stage, and Resourceshows the mineral resources each mining city reserves and pro-duces. As a result, mining city type dimension is needed. Conse-quently, there are eight dimensions in the ISMCS altogether.

Furthermore, since association rule mining is an unsupervisedlearning method, there is no clear measure of success, which canbe used to judge adequacy in indicator selection (Hastie et al.,2009). Therefore, to effectively discover the novel, implicit asso-ciation rules of mining city sustainability, we select relatively lar-ger number of indicators than previous research (Yu et al., 2005;Yu et al., 2008), so as to collect adequate data on mining citysustainability. In consequence, we constructed the ISMCS usingeight dimensions and 52 indicators (Table 1).

2.2. Sample selection and data collection

Considering comprehensiveness and comparability for ourresearch sample, we chose all prefecture-level mining cities

Table 1Indicator System for Mining City Sustainability.

Dimension Indicator

Economy I11: GDP,I12: GDP per capita,I13: Growth rate of GDP,I14: Total fixed assetI15: Proportion of third industry value added toGDP,I16: Growth rate of Investment in fixed assetI17: Household saving deposits per capita

Society I21: Total population at year-endI22: Annual growth rate of populationI23: Number of books in public library per hundredpopulationsI24: Hospital doctors per ten thousand populationsI25: Number of beds of hospital and health centerper ten thousands of populationsI26: Number of subscribers of mobile telephoneper hundred populationsI27: Number of subscribers of Internet services perhundred populationsI28: Proportion of expenditure of science andtechnology to local government general budgetexpenditureI29: Proportion of expenditure of education to localgovernment general budget expenditure

Environment I31: Number of days when air quality reached thegrade-two of the national standardI32: Ratio of waste water centralized treated insewage workI33: Ratio of consumption wastes treatedI34: Ratio of green covered area of completed areaI35: Area of green land per capita

Economy-society I41: Job creation for all sectors combinedI42: Unemployment rateI43: Proportion of mining population to totalworking populationI44: Fixed asset per capitaI45: Total amount of profit and tax of industrialenterprises above designated sizeI46: Profit and tax contributed from per hundredyuan of fixed assetsI47: Population receiving subsistence allowanceI48: Proportion of investment on education to GDPI49: Proportion of investment on science andtechnology to GDP

Society-environment I51: Annual water consumptionI52: Annual electricity consumptionI53: Daily domestic water consumption per personI54: Daily household electricity consumption perperson

Economy-environment I61: Electricity consumption per yuan of GDPI62: Water consumption per yuan of GDPI63: Volume of industrial waste water dischargedI64: Volume of industry sulphur dioxide emissionI65: Ratio of general industrial solid wastes com-prehensively utilizedI66: Volume of industrial dust emissionI67: Volume of industrial waste water dischargedper ten thousand yuan of GDPI68: Volume of industry sulphur dioxide emissionper ten thousand yuan of GDPI69: Volume of industrial dust emission per tenthousand yuan of GDP

Economy-society-environment

I71: Number of public transportation vehicles per10,000 populationsI72: Area of paved road per capitaI73: Daily water consumption per personI74: Daily electricity consumption per personI75: Population density of municipal district

Mining city Type I81: ProvinceI82: RegionI83: StageI84: Resource

Table 2Basic data of 110 prefecture-level mining cites in China.

Region Province City name Stage Resource

North China Hebei Zhangjiakou Mature CoalChengde Mature FerrousTangshan Reborn CoalHandan Mature CoalXingtai Mature Coal

Shanxi Datong Mature CoalYangquan Mature CoalChangzhi Mature CoalJincheng Mature CoalShuozhou Growing CoalJinzhong Mature CoalYuncheng Mature CoalXinzhou Mature CoalLinfen Mature CoalLvliang Mature Coal

InnerMongolia

Baotou Reborn Multi-resources

Wuhai Declining CoalChifeng Mature CoalOrdos Growing CoalHulun Buir Growing Coal

NortheastChina

Liaoning Anshan Reborn FerrousFushun Declining CoalBenxi Mature FerrousFuxin Declining CoalPanjin Reborn PetroleumHuludao Reborn Nonferrous

Metal

Jilin Songyuan Growing PetroleumLiaoyuan Declining CoalTonghua Reborn Coal

Heilongjiang Jixi Mature CoalHegang Declining CoalShuangyashan Declining CoalDaqing Mature PetroleumQitaihe Declining Coal

East China Jiangsu Xuzhou Reborn CoalSuqian Reborn Other

Zhejiang Huzhou Mature Other

Anhui Huainan Mature CoalMa'anshan Reborn FerrousHuaibei Declining CoalTongling Declining Nonferrous

MetalChuzhou Mature OtherSuzhou Mature CoalBozhou Mature CoalChizhou Mature OtherXuancheng Mature Other

Fujian Sanming Mature Multi-resources

Nanping Mature NonferrousMetal

Longyan Mature Multi-resources

Jiangxi Jingdezhen Declining OtherXinyu Declining FerrousPingxiang Declining CoalGanzhou Mature Nonferrous

Metal

L. Zeng et al. / Resources Policy 49 (2016) 394–404396

Table 2 (continued )

Region Province City name Stage Resource

Yichun Mature Multi-resources

Shandong Zibo Reborn CoalZaozhuang Declining CoalDongying Mature PetroleumJining Mature CoalTai'an Mature CoalLaiwu Mature FerrousLinyi Reborn Coal

Central China Henan Luoyang Reborn NonferrousMetal

Pingdingshan Mature CoalHebi Mature CoalJiaozuo Declining CoalPuyang Declining PetroleumSanmengxia Mature Nonferrous

MetalNanyang Reborn Petroleum

Hubei Ezhou Mature FerrousHuangshi Declining Coal

Hunan Hengyang Mature Multi-resources

Chenzhou Mature FerrousShaoyang Mature CoalLoudi Mature Multi-

resources

South China Guangdong Shaoguan Declining FerrousYunfu Mature Other

Guangxi Baise Mature NonferrousMetal

Hezhou Growing NonferrousMetal

Hechi Mature NonferrousMetal

SouthwestChina

Sichuan Zigong Mature OtherPanzhihua Mature FerrousLuzhou Declining Multi-

resourcesGuangyuan Mature CoalNanchong Growing Multi-

resourcesGuang'an Mature CoalDazhou Mature CoalYa'an Mature Multi-

resources

Guizhou Liupanshui Growing CoalAnshun Mature CoalBijie Growing Coal

Yunnan Qujing Mature CoalBaoshan Mature Multi-

resourcesZhaotong Growing CoalPu'er Mature Nonferrous

MetalLincang Mature Nonferrous

Metal

NorthwestChina

Shaanxi Tongchuan Declining CoalBaoji Mature Nonferrous

Metal

Table 2 (continued )

Region Province City name Stage Resource

Xianyang Growing CoalWeinan Mature Nonferrous

MetalYan'an Growing PetroleumYulin Growing Coal

Gansu Jinchang Mature NonferrousMetal

Baiyin Declining NonferrousMetal

Wuwei Growing CoalZhangye Reborn FerrousPingliang Mature CoalQingyang Growing PetroleumLongnan Growing Nonferrous

MetalNingxia Shizuishan Declining CoalXinjiang Kelamayi Mature Petroleum

L. Zeng et al. / Resources Policy 49 (2016) 394–404 397

identified by the State Council of China (2013) (The State Councilof the People's Republic of China, 2013). Among these cities, somedata from 9 autonomous prefectures were not available. Conse-quently, we selected 110 prefecture-level cities as research cities(Table 2) and 2012 as research period. The stage type and provinceof mining cities were determined according to the State Council ofChina, and the resource type of mining cities was determinedaccording to the Research Group from Macroeconomic ResearchAcademy of the State Planning Commission and Mineral ResourcePlanning of every mining city. Other data were gotten from 2013China City Statistical Yearbook, 2012 China City Statistical Yearbook,2013 Statistical Yearbook and 2012 Statistical Bulletin of somemining cities.

The boundaries of administrative units were based on the Na-tional Geomatics Center of China at the scale of 1:4,000,000. Ourstudy focused only on the mainland China although Taiwan, HongKong, and Macao were still present in the figures.

2.3. Association rule mining principles and algorithm

Association rule mining is one of the most popular data-miningmethods. Data mining, which is also popularly referred to asKnowledge Discovery from Data (KDD), is the process of dis-covering interesting knowledge from large amounts of data storedin databases, data warehouses, or other information repositories.Association rule mining leads to the discovery of frequent patterns,associations, and many other interesting relationships among datasources. A typical example of Association rule mining is marketbasket analysis. By finding which items are frequently placed to-gether in customer's supermarket shopping baskets, such as milkand bread, Association rule mining can discover which items arefrequently purchased together by customers (frequent pattern),and gain insight into their customers' buying habits. For instance,if customers are buying milk, how likely are they to also buy breadin the same trip to the supermarket (association rule)? This kind ofmining and analysis can help retailers develop marketing strate-gies, such as selective marketing and planning shelf space (Hanet al., 2011).

Association rule mining is very effective in identifying frequentpatterns and interesting relationships that are novel, implicit, andpreviously unknown. Frequent patterns, as the name suggests, arepatterns that occur frequently in data. By performing Associationrule mining on a Chinese mining cities sustainability dataset, wecan identify the frequent patterns in the dataset. Therefore, bydiscovering the sustainability patterns of these cities, we can also

L. Zeng et al. / Resources Policy 49 (2016) 394–404398

uncover the novel, implicit and previously unknown relationshipsin the dataset and the sustainability characteristics of these miningcities. These results can be the basis of policy making.

2.3.1. Notations and definitionsFormal notations and definitions are given as follows: set of

items: I ¼ {I1, I2, : : : , Im}; set of transactions: D, the task-relevantdataset, each transaction T is a set of items such that T D I; eachtransaction is associated with an identifier, called TID. Let A be aset of items. A transaction T is said to contain A if and only if A D T.

Support of A is the percentage of transactions in D that containA. An itemset that contains k items is a k-itemset. If the support ofan itemset I satisfies a prespecified minimum support threshold,then I is a frequent itemset. The set of frequent k-itemset iscommonly denoted by Lk. An association rule is an implication ofthe form:

⇒ ( )⎡⎣ ⎤⎦A B support confidence, 2.1

where A ⊂ I, B ⊂ I , and A ∩ B ¼ Φ; support and confidence aremeasures to evaluate an association rule (Agrawal et al., 1993;Agrawal and Srikant, 1994; Kotsiantis and Kanellopoulos, 2006;Han et al., 2011; Chae et al., 2001; Tsay and Chiang, 2005).

2.3.2. Measures to evaluate association rulesA data mining system has the potential to generate thousands

or even millions of patterns, or rules. However, only a small part ofthe patterns would be interesting. There are some objectivemeasures and subjective measures to evaluate pattern interest-ingness. A pattern is interesting if it is: (1) easily understood byhumans, (2) valid on new or test data with some degree of cer-tainty, (3) potentially useful, and (4) novel. A pattern is also in-teresting if it validates a hypothesis that the user sought. These aresubjective measures of pattern interestingness. There are stillseveral objective measures that are based on the structure ofdiscovered patterns and the statistics underlying them. For asso-ciation rules, there are three important objective measures: sup-port, confidence, and lift.

2.3.2.1. Support. The rule A ⇒ B holds in the transaction set D withsupport s, where s is the percentage of transactions in D thatcontain A ∪ B (i.e., the union of sets A and B, or say, both A and B).This is taken to be the probability, P (A ∪ B). That is:

( ⇒ ) = ( ∪ ) ( )Support A B P A B 2.2

2.3.2.2. Confidence. The rule A ⇒ B has confidence c in the trans-action set D, where c is the percentage of transactions in D con-taining A that also contain B. This is taken to be the conditionalprobability, P (B|A). That is:

( )( ⇒ ) = | ( )Confidence A B P B A 2.3

Rules that satisfy both a minimum support threshold (min_sup)and a minimum confidence threshold (min_conf) are called strong(Agrawal et al., 1993; Agrawal and Srikant, 1994; Kotsiantis andKanellopoulos, 2006; Han et al., 2011; Chae et al., 2001; Tsay andChiang, 2005).

2.3.2.3. Lift. Most Association rule mining algorithms employ asupport-confidence framework. Although minimum support andconfidence thresholds help weed out or exclude the exploration ofa good number of uninteresting rules, many rules thus generatedare still not interesting to the users.

To tackle this weakness, lift can be used to augment the sup-port-confidence framework for association rules. Lift is a simplecorrelation measure that is given as follows. The occurrence of

itemset A is independent of the occurrence of itemset B if P (A∪B)¼ P(A)P(B); otherwise, itemsets A and B are dependent and cor-related as events.

( )( ) ( ) ( )= ( ∪ ) ( )lift A B P A B P A P B, / 2.4

If the resulting value of Eq. (2.4) is less than 1, then the oc-currence of A is negatively correlated with the occurrence of B. Ifthe resulting value is greater than 1, then A and B are positivelycorrelated, meaning that the occurrence of one implies the oc-currence of the other. If the resulting value is equal to 1, then A andB are independent and there is no correlation between them (Hanet al., 2011).

2.3.3. Association rule mining processThe occurrence frequency of an itemset is the number of

transactions that contain the itemset. This is also known as thefrequency or support count. Note that the itemset support definedin Eq. (2.2) is sometimes referred to as relative support, whereasthe occurrence frequency is called the absolute support. If therelative support of an itemset I satisfies a prespecified minimumsupport threshold (i.e., the absolute support of I satisfies the cor-responding minimum support count threshold), then I is a fre-quent itemset. From Eq. (2.3), we have:

( )( )

( ⇒ ) = | = _

( ∪ ) _ ( )

Confidence A B P B A support

count A B support count A/ 2.5

Eq. (2.5) shows that the confidence of rule A ⇒ B can be easilyderived from the support count of A and A ∪ B. That is, once thesupport count of A, B, and A∪B are found, it is straightforward toderive the corresponding association rules A ⇒ B and B ⇒ A andcheck whether they are strong. Thus, the problem of mining as-sociation rules can be reduced to that of mining frequent itemsets.

In general, association rule mining can be viewed as a two-stepprocess: (1) find all frequent itemsets. By definition, each of theseitemsets will occur at least as frequently as a predeterminedminimum support count, min_sup. (2) Generate strong associationrules from the frequent itemsets: By definition, these rules mustsatisfy minimum support and minimum confidence. Because thesecond step is much less costly than the first, the overall perfor-mance of mining association rules is determined by the first step(Han et al., 2011).

2.3.4. The apriori algorithmApriori algorithm is the most popular algorithm for mining

association rules. Apriori employs an iterative approach known asa level-wise search, where k-itemsets are used to explore (kþ1)-itemsets. First, by scanning the database to accumulate the countfor each item and collecting those items that satisfy minimumsupport, the set of frequent 1-itemsets is found. The resulting set isdenoted L1.Next, L1 is used to find L2, the set of frequent 2-item-sets, which is used to find L3, and so on, until no more frequent k-itemsets can be found. The finding of each Lk requires one full scanof the database.

To improve the efficiency of the level-wise generation of fre-quent itemsets, an important property called the apriori propertyis used to reduce the search space, and the apriori property is thatall nonempty subsets of a frequent itemset must also be frequent(Agrawal et al., 1993;Han et al., 2011).

2.4. Association rule mining implementation

We used Clementine 12.0 professional data mining software asa platform to implement association rule mining. Before mining,some preprocessing methods were applied on the collected

L. Zeng et al. / Resources Policy 49 (2016) 394–404 399

Chinese mining city sustainability dataset to improve the dataquality. The mean value of an indicator was used to fill in the fewmissing values, quartile was used to discretize the continuous in-dicators, and then the preprocessed dataset were generated. TheApriori model was applied to the preprocessed dataset to mineassociation rules.

In general, this study employs these steps to mine associationrules: dataset selection, antecedents and consequents specifica-tion, setting support and confidence threshold, mining associationrules, rules evaluation, and rules presentation and explanation. Inmining valuable association rules, these steps need to be repeatedwith different antecedents and consequents and different supportand confidence thresholds.

To effectively discover the association rules between differentsustainability indicators and identify the development patterns ofChinese mining cities, the implementation of association rulesmining was organized as follows. First, we selected all the in-dicators as antecedent and consequent to mine the associationrules between two indicators. Then, we selected all the indicatorsas antecedents and consequents to mine the association rulesbetween multiple indicators. Finally, we selected mining city typeindicators as antecedent and select all the indicators as con-sequent to mine association rules about different types of miningcities.

Table 3Association rules between two indicators.

Rule ID Association Rules

Antecedent ⇒ Co

3-1 Volume of industrial dust emission per ten thousand yuan of GDP ¼high

⇒ Re

3-2 Proportion of mining population to total working population ¼ high ⇒ Re3-3 Proportion of investment on education to GDP ¼ low ⇒ G3-4 GDP per capita ¼ high ⇒ Pr3-5 GDP per capita ¼ low ⇒ Pr3-6 Proportion of investment on education to GDP ¼ high ⇒ G3-7 Household saving deposits per capita ¼ low ⇒ Pr3-8 Number of subscribers of Internet services per hundred population ¼

low⇒ G

3-9 GDP per capita ¼ low ⇒ Nlo

3-10 Number of subscribers of mobile telephone per hundred population¼ low

⇒ G

3-11 GDP per capita ¼ low ⇒ Nlo

3-12 Area of green land per capita ¼ low ⇒ G3-13 GDP per capita ¼ low ⇒ A3-14 Number of subscribers of Internet services per hundred population ¼

low⇒ H

3-15 Household saving deposits per capita ¼ low ⇒ Nlo

3-16 Number of subscribers of mobile telephone per hundred population¼ low

⇒ Npo

3-17 Number of beds of hospital and health center per ten thousands ofpopulation ¼ low

⇒ Nlo

3-18 Household saving deposits per capita ¼ low ⇒ H3-19 Hospital doctors per ten thousand population ¼ low ⇒ H3-20 GDP per capita ¼ high ⇒ N3-21 Number of books in public library per hundred population ¼ high ⇒ G3-22 GDP per capita ¼ high ⇒ N

hi3-23 Number of subscribers of mobile telephone per hundred population

¼ high⇒ G

3-24 Number of subscribers of Internet services per hundred population ¼high

⇒ H

3-25 Household saving deposits per capita ¼ high ⇒ Nhi

Setting: Minimum antecedent support ¼ 10%, Minimum rule confidence ¼ 70%, MaximResult summary: Number of Rules: 71, Number of Valid Transactions: 110, Minimum SConfidence: 90.9%.

3. Results

First, we applied the mining city sustainability dataset to mineall the association rules between two indicators to produce aholistic and clear knowledge about the sustainability of miningcities in China. The minimum antecedent support and minimumrule confidence employed here are 10% and 70%, respectively. Theminimum lift is automatically determined by software as morethan 1. Seventy-one association rules were generated. The sum-mary of the results are as follows: minimum support, 10.0%;maximum support, 25.5%; minimum confidence, 70.0%; maximumconfidence, 93.8%; minimum lift, 1.3; and maximum lift, 3.7. Afterremoving repeated, meaningless, and incomprehensible rules, 33rules remained (Table 3 and Fig. 2). In Tables 3–5, C means ruleconfidence and unit is %, S means rule support and unit is %.

Then, we continued to apply the dataset to mine all the asso-ciation rules between multiple indicators to produce a holistic andthorough knowledge about the sustainability of mining cities inChina. The minimum antecedent support and minimum ruleconfidence employed here are 15% and 80%, respectively. Onehundred and sixty-three association rules were generated. Thesummary of the results are as follows: minimum support, 15.5%;maximum support, 25.5%; minimum confidence, 80.0%; maximumconfidence, 100%; minimum lift, 1.5; and maximum lift, 3.7. After

C % S%

nsequent

source ¼ Coal 82.1 20.9

source ¼ Coal 82.1 20.9DP per capita ¼ high 77.8 19.1oportion of investment on education to GDP ¼ low 75.0 19.1oportion of investment on education to GDP ¼ high 74.1 18.2DP per capita ¼ low 71.4 18.2oportion of investment on education to GDP ¼ high 74.1 18.2DP per capita ¼ low 74.1 18.2

umber of subscribers of Internet services per hundred population ¼w

74.1 18.2

DP per capita ¼ low 70.4 17.3

umber of subscribers of mobile telephone per hundred population ¼w

70.4 17.3

DP per capita ¼ low 70.4 17.3rea of green land per capita ¼ low 70.4 17.3ousehold saving deposits per capita ¼ low 77.8 19.1

umber of subscribers of Internet services per hundred population ¼w

77.8 19.1

umber of beds of hospital and health center per ten thousands ofpulation ¼ low

74.1 18.2

umber of subscribers of mobile telephone per hundred population ¼w

74.1 18.2

ospital doctors per ten thousand population ¼ low 74.1 18.2ousehold saving deposits per capita ¼ low 74.1 18.2umber of books in public library per hundred population ¼ high 75.0 19.1DP per capita ¼ high 75.0 19.1umber of subscribers of mobile telephone per hundred population ¼gh

71.4 18.2

DP per capita ¼ high 71.4 18.2

ousehold saving deposits per capita ¼ high 71.4 18.2

umber of subscribers of Internet services per hundred population ¼gh

71.4 18.2

um number of antecedents ¼ 1.upport: 10%, Maximum Support: 25. 5%, Minimum Confidence: 70.0%, Maximum

Fig. 2. Support and confidence of Association rules between two indicators.

L. Zeng et al. / Resources Policy 49 (2016) 394–404400

removing repeated, meaningless, and incomprehensible rules, 21rules remain (Table 4 and Fig. 3).

Finally, to find specific association rules related to familiarclassification indicators for mining cities, such as developmentstage, resource type, and region, and reveal the sustainabilitycharacteristics and even patterns of different mining city types,and make corresponding policies for different types of mining ci-ties, we employed the dataset to find association rules with thethree antecedents. Here we selected stage, region and resource asantecedents, and choose all indicators as consequents. We setminimum antecedent support and minimum rule confidenceemployed as 10% and 70%, respectively. Twenty-one associationrules were generated. The summary of the results are as follows:minimum support, 10%; maximum support, 18.2%; minimumconfidence, 70.0%; maximum confidence, 92.9%; minimum lift,1.28; and maximum lift, 3.85. After removing three repeated rules,there were still 18 rules (Table 5 and Fig. 4).

4. Discussion

The mined association rules are summaries of implicit rulesthat underlie the dataset. These rules reflect the patterns andcharacteristics of mining city sustainability in China.

4.1. Association rules between two indicators

As we can see from Table 3, there are some distinct char-acteristics and notable patterns on mining city sustainability inChina.

Rule 3-1 indicates that 82.1% of the mining cities with a highvalue of volume of industrial dust emission per 10,000 yuan ofGDP are coal-mining cities. Rule 3-2 shows that 82.1% of themining cities with a high proportion of mining population to totalworking population are also coal-mining cities. These two rulestell us that coal-mining cities in China have distinctive character-istics in that they often have a larger proportion of mining popu-lation and greater pressure on reducing industrial dust emission.

Rules 3-3 to 3-7 are novel association rules about mining citysustainability in China. Rules 3-3 and 3-4 show strong associationsbetween low proportion of investment in education to GDP andhigh GDP per capita; in contrast, Rules 3-5 and 3-6 show strongassociations between high proportion of investment in educationto GDP and low GDP per capita. Rule 3-7 reveals the strong asso-ciation between a low value of Household saving deposits percapita and high proportion of investment in education in miningcities in China. These five rules show the extreme imbalance be-tween the economy and education in mining cities in China.

Rule 3-8 and Rule 3-9 show the associations between low GDPper capita and low number of subscribers to Internet services per100 populations. Rules 3-10 and 3-11 indicate associations

between low GDP per capita and low number of mobile telephonesubscribers per 100 populations. Rules 3-12 and 3-13 show asso-ciations between low GDP per capita and low area of green landper capita. The above six rules mean that a low level of economicdevelopment is often accompanied by low levels of social devel-opment and environmental construction in mining cities in China.

Rules 3-14 and 3-15 show the associations between the lownumber of subscribers to Internet services per 100 populationsand low household saving deposits per capita. Rules 3-16 and 3-17present the associations between the low number of mobile tel-ephone subscribers per 100 populations and low number of hos-pital and health center beds per 10,000 of population. Rule 3-18and Rule 3-19 discover the associations between the low house-hold saving deposits per capita and low number of hospital doc-tors per 10,000 populations. These six rules indicate that low le-vels of social development in some areas often coexist in somemining cities in China.

Rules 3-20 and 3-21 show the associations between high GDPper capita and high number of books in public libraries per 100populations. Rules 3-22 and 3-23 reveal the associations betweenhigh GDP per capita and high number of mobile telephone sub-scribers per 100 populations. Rules 3-24 and 3-25 reveal the as-sociations between a high number of subscribers to Internet ser-vices per 100 populations and high household saving deposits percapita. These six rules tell us that a high level of economic de-velopment is often accompanied by a high level of social devel-opment in some aspects of mining cities in China, and high levelsof social development often coexist in a few fields. Nevertheless,the coexistence effect of low levels of different dimensions ofsustainability is much stronger than that of high levels of differentdimensions of sustainability.

By analyzing the association rules between two indicators, weidentify some characteristics of mining city sustainability in China.Coal-mining cities usually have a larger proportion of the miningpopulation and should pay greater attention to reducing industrialdust emissions than other kinds of mining cities. Investment ineducation, economic development, and other aspects of society areextremely unbalanced. A high level of economic development isoften accompanied by a high level of social development and en-vironmental construction in some mining cities. At the same time,a low level of economic development, social development andenvironmental construction coexist much more often in somemining cities.

4.2. Association rules between multiple indicators

Table 4 shows that there are some thorough and complicatedassociations between multiple indicators of mining cities in China.

Rules 4-1, 4-2, and 4-3 are about the associations between thelow proportion of investment in education, high area of green landper capita, high number of books in public libraries per 100 po-pulations, and high GDP per capita. In contrast, Rules 4-4 and 4-5show that a high proportion of investment in education to GDPhas an association with low area of green land per capita, house-hold saving deposits per capita, and GDP per capita. The above fiverules further demonstrate the imbalance between investment ineducation, economic development, and some aspects of societyand environment protection in mining cities in China.

Rules 4-6, 4-7 and 4-8 show associations between a high areaof green land, high number of books in public libraries per 100populations, and high GDP per capita. Rule 4-9 presents the as-sociation between a high number of mobile telephone subscribersper 100 populations, high number of books in public libraries per100 population, and high GDP per capita. These four rules revealthat a high level of economic development is often accompaniedby high levels of specific aspects of social development and

Table 4Association rules between multiple indicators.

Rule ID Association Rules C % S %

Antecedent ⇒ Consequent

4-1 Proportion of investment on education to GDP ¼ low and Area of green land per capita ¼ high ⇒ GDP per capita ¼ high 100 15.54-2 Proportion of investment on education to GDP ¼ low and GDP per capita ¼ high ⇒ Area of green land per capita ¼ high 81.0 15.54-3 Proportion of investment on education to GDP ¼ low and Number of books in public library per hundred population ¼ high ⇒ GDP per capita ¼ high 94.1 14.54-4 Area of green land per capita ¼ low and Household saving deposits per capita ¼ low ⇒ Proportion of investment on education to GDP ¼ high 88.2 13.64-5 Household saving deposits per capita ¼ low and GDP per capita ¼ low ⇒ Proportion of investment on education to GDP ¼ high 88.9 14.54-6 Area of green land per capita ¼ high and Number of books in public library per hundred population ¼ high ⇒ GDP per capita ¼ high 94.4 15.54-7 GDP per capita ¼ high and Area of green land per capita ¼ high ⇒ Number of books in public library per hundred population ¼ high 89.5 15.54-8 GDP per capita ¼ high and Number of books in public library per hundred population ¼ high ⇒ Area of green land per capita ¼ high 81.0 15.54-9 Number of subscribers of mobile telephone per hundred population ¼ high and Number of books in public library per hundred

population ¼ high⇒ GDP per capita ¼ high 88.9 14.5

4-10 Number of subscribers of mobile telephone per hundred population ¼ low and Number of subscribers of Internet services perhundred population ¼ low

⇒ GDP per capita ¼ low 88.9 14.5

4-11 Number of subscribers of mobile telephone per hundred population ¼ low and GDP per capita ¼ low ⇒ Number of subscribers of Internet services per hundred population¼ low

84.2 14.5

4-12 Household saving deposits per capita ¼ low and Number of subscribers of Internet services per hundred population ¼ low ⇒ GDP per capita ¼ low 81.0 15.54-13 Household saving deposits per capita ¼ low and GDP per capita ¼ low ⇒ Number of subscribers of Internet services per hundred population

¼ low94.4 15.5

4-14 Area of green land per capita ¼ low and Household saving deposits per capita ¼ low ⇒ Number of subscribers of Internet services per hundred population¼ low

94.1 14.5

4-15 Area of green land per capita ¼ low and Number of subscribers of Internet services per hundred population ¼ low ⇒ Household saving deposits per capita ¼ low 88.9 14.54-16 Area of green land per capita ¼ low and Hospital doctors per ten thousand population ¼ low ⇒ Number of subscribers of Internet services per hundred population

¼ low88.2 13.6

4-17 Area of green land per capita ¼ low and Hospital doctors per ten thousand population ¼ low ⇒ Household saving deposits per capita ¼ low 88.2 13.64-18 Area of green land per capita ¼ low and Household saving deposits per capita ¼ low ⇒ Hospital doctors per ten thousand population ¼ low 88.2 13.64-19 Number of subscribers of Internet services per hundred population ¼ low and Hospital doctors per ten thousand population ¼

low⇒ Household saving deposits per capita ¼ low 94.4 15.5

4-20 Hospital doctors per ten thousand population ¼ low and Household saving deposits per capita ¼ low ⇒ Number of subscribers of Internet services per hundred population¼ low

85.0 15.5

4-21 Number of subscribers of Internet services per hundred population ¼ low and Household saving deposits per capita ¼ low ⇒ Hospital doctors per ten thousand population ¼ low 81.0 15.5

Setting: Minimum antecedent support ¼ 15%, Minimum rule confidence ¼ 80%, Maximum number of antecedents ¼ 5.Result summary: Number of Rules: 163, Number of Valid Transactions: 110, Minimum Support: 15.5%, Maximum Support: 25.5%, Minimum Confidence: 80.0%, Maximum Confidence: 100.0%.

L.Zenget

al./Resources

Policy49

(2016)394

–404401

Table 5Association rules with specified antecedents.

Rule ID Association Rules C % S%

Antecedent ⇒ Consequent

5-1 Region ¼ Northeast China ⇒ Annual growth rate of population ¼ low 85.7 10.95-2 Region ¼ Northeast China ⇒ Number of beds of hospital and health center per ten thousands of population ¼

high71.4 9.1

5-3 Region ¼ East China and Resource ¼ Coal ⇒ Population density of municipal district ¼ medium-high 72.7 7.35-4 Region ¼ East China and Resource ¼ Coal ⇒ Volume of industrial waste water discharged per ten thousand yuan of GDP ¼

low-medium72.7 7.3

5-5 Region ¼ Southwest China ⇒ Growth rate of GDP ¼ high 87.5 12.75-6 Region ¼ Southwest China ⇒ Household saving deposits per capita ¼ low 81.3 11.85-7 Region ¼ Southwest China and Stage ¼ Mature ⇒ Growth rate of GDP ¼ high 81.8 8.25-8 Region ¼ Southwest China and Stage ¼ Mature ⇒ GDP per capita ¼ low 72.7 7.35-9 Region ¼ Southwest China and Stage ¼ Mature ⇒ Number of days when air quality reached the grade-two of the national standard

¼ high72.7 7.3

5-10 Region ¼ Southwest China and Stage ¼ Mature ⇒ Hospital doctors per ten thousand population ¼ low 72.7 7.35-11 Region ¼ North China ⇒ Resource ¼ Coal 90.0 16.45-12 Region ¼ North China ⇒ Stage ¼ Mature 70.0 12.75-13 Region ¼ North China ⇒ Volume of industrial dust emission ¼ high 70.0 12.75-14 Region ¼ North China and Stage ¼ Mature ⇒ Volume of industrial dust emission ¼ high 71.4 9.15-15 Region ¼ North China and Resource ¼ Coal ⇒ Volume of industrial dust emission ¼ high 72.2 11.85-16 Region ¼ North China and Resource ¼ Coal and Stage ¼

Mature⇒ Volume of industrial dust emission ¼ high 76.9 9.1

5-17 Region ¼ North China and Stage ¼ Mature ⇒ Growth rate of GDP ¼ low 71.4 9.15-18 Resource ¼ Ferrous ⇒ Ratio of consumption wastes treated ¼ medium-high 90.9 9.1

Setting: Minimum antecedent support ¼ 10%, Minimum rule confidence ¼ 70%, Maximum number of antecedents ¼ 5.Result summary: Number of Rules: 21, Number of Valid Transactions: 110, Minimum Support: 10%, Maximum Support: 18.2%, Minimum Confidence: 70.0%, MaximumConfidence: 92.9%.

Fig. 3. Support and confidence of Association rules between multiple indicators.

Confidence (%) Support (%)

Fig. 4. Support and confidence of Association rules with specified antecedents.

L. Zeng et al. / Resources Policy 49 (2016) 394–404402

greening construction in mining cities in China.Rules 4-10 to 4-13 find that a low level of economic develop-

ment often coexists with a low level of development of multipleaspects of society in mining cities in China. Rules 4-14 to 4-18 notethat a low level of greening construction is usually accompaniedby low levels of multiple fields of social development. Additionally,Rules 4-19 to 4-21 show that of a low level of social development

in multiple fields often exist at the same time in some miningcities in China.

Therefore, it can be seen that the association rules betweenmultiple indicators further demonstrate the imbalanced situationbetween investment in education and some other fields of sus-tainable development of mining cities and the coexistence effect oflow levels of different sustainability dimensions is much strongercompared with high levels of different sustainability dimensions.

4.3. Association rules with specified antecedents

Table 5 consists of 18 new and valuable association rules minedwith the three specified antecedents: Region, Stage, and Resource.These rules present some distinct characteristics and notablepatterns of the sustainability of some types of mining cities inChina. In the 18 mined rules, 17 are about region, 8 are aboutdevelopment stage, and 6 are about resource type.

Rule 5-1 shows that 85.7% of mining cities in northeast Chinahave a low value of annual growth rate of population. Additionally,Rule 5-2 notes that 71.4% of mining cities have a high number ofhospital and health center beds per 10,000 of populations innortheast China. The two rules reveal two characteristics of socialdevelopment in most mining cities in northeast China: a low po-pulation growth rate and good quality of medical facilities.

Rules 5-3 and 5-4 are two rules about coal-mining cities in eastChina that show that 72.7% of coal-mining cities in east China havea medium-high municipal district population density and low-medium volume of industrial wastewater discharged per 10,000yuan of GDP.

Rules 5-5 to 5-10 show some sustainability characteristics ofmining cities in southwest China. Rule 5-3 tells us that 87.5% ofmining cities in southwest China have a high GDP growth rate;however, Rule 5-6 shows that 81.3% of mining cities in southwestChina have low household saving deposits per capita. Rules 5-7 to5-10 are four rules about mature mining cities in southwest China.Rule 5-7 indicates that 81.8% of these cities have a high GDPgrowth rate; however, Rule 5-8 shows that 72.7% of these cities

Fig. 5. Distinctive regional characteristics of mining city sustainability in China.

L. Zeng et al. / Resources Policy 49 (2016) 394–404 403

have a low GDP per capita. Rules 5-9 and 5-10 show that 72.7% ofmature mining cites in southwest China have a high number ofdays when air quality reached grade two of the national standardand a low number of hospital doctors per 10,000 population.

Rules 5-11 to 5-17 are seven rules about mining cities in northChina. Rules 5-10 to 5-12 indicate that in the north Chinese miningcities, 90% are coal-mining cities and 70% are mature mining cities.Rules 5-13 to 5-16 show that north Chinese mining cities experi-ence bad level of industrial dust emission, which means that 70%of mining cities, 71.4% of mature mining cities, 72.2% of coal-mining cities, and 76.9% of mature coal-mining cities have a highvolume of industrial dust emissions. Rules 5-17 tells us that 70% ofmature mining cities in north China have a low GDP growth rate.

Rule 5-18 shows that 90.9% of ferrous mining cities have amedium–high ratio of treated consumption waste.

From the above rules, we learn that most mining cities innortheast China have a low population growth rate and goodquality medical facilities. In most east Chinese coal-mining cities,the municipal district population density is medium-high and in-dustrial wastewater is not fully discharged. Most mining cities insouthwest China have a high GDP growth rate and good air quality,but they still have a low level of economic and social development.Most mining cities in north China are mature mining cities andcoal-mining cities; at the same time, most north Chinese miningcities experience bad level of industrial dust emissions, and mostmature mining cities in north China have a low GDP growth rate. It

is clear that mining city sustainability in China presents distinctiveregional characteristics. We can also see that mining city sustain-ability characteristics and patterns in China have much to do withtheir regions and have something to do with their socioeconomicdevelopment stage and primary mineral resources (Fig. 5).

Our results suggest that different region of China have com-pletely different landscape patterns and numerous studies in thefield of landscape ecology have shown that patterns of urbanlandscapes have substantial influences on ecological and socio-economic process (Huang et al., 2015). As a result, we should paygreater attention to regional characteristics and differences whenmaking sustainability planning and policies than the other twokinds of differences, development stage or resources type, topromote mining city sustainability in China. In fact, in the na-tionwide resource-based cities sustainable development planning(2013–2020), too much importance has been given to the devel-opment stage and not to regional difference. Especially, to promotemining city sustainability more scientifically, spatial heterogeneitywithin a mining city and broader region needs to be considered(Wu, 2010).

5. Conclusions

Mining cities are of vital importance to the socioeconomic de-velopment of China during recent decades, but there are great

L. Zeng et al. / Resources Policy 49 (2016) 394–404404

challenges for their sustainability. Understanding the characteristicsand patterns of mining city sustainability is critical to promotingtheir sustainability. After constructing the ISMCS and collecting datafrom all available 110 prefecture-level mining cities, we applied theAssociation rule mining method to the collected dataset and ac-quired 64 interesting association rules. Through analyzing theseassociation rules and corresponding frequent patterns, we identi-fied some novel, implicit, and previously unknown characteristicsand patterns of mining city sustainability in China.

First, investment in education, economic development, andsome aspects of society are extremely unbalanced in most miningcities. On one hand, a high level of economic development is oftenaccompanied by a high level of social development and environ-mental construction in some mining cities. On the other hand, alow level of socioeconomic development and environmentalconstruction much more often coexist in some mining cities.

Second, most coal-mining cites have larger proportion of themining population and usually face severe challenges in reducingindustrial dust emissions.

Last but not least, mining city sustainability in China exhibitsdistinctive regional characteristics. Most mining cities in northeastChina have a low population growth rate and good quality medicalfacilities. Most mining cities in southwest China have a high GDPgrowth rate and good air quality, but they usually have a low levelof socioeconomic development. And most mining cities in northChina face a serious situation in reducing industrial dust emissionsto improve air quality. Landscape pattern analysis should be in-troduced to scientifically promote mining city sustainability.

These novel and previously unknown mining city sustainabilityin China fully demonstrate the effectiveness of Association rulemining in sustainability analysis and are expected to provide abasis for decision-making for of mining city sustainability in China.

Funding

This research was sponsored by National Natural ScienceFoundation of China (Grant No. 71303184) and Shandong ProvinceSoft Science Research Program (Grant No. 2014RKB01546).

Conflicts of interest

The authors declare no conflict of interest.

Author contributions

Lijun Zeng designed and carried out the research and draftedthe paper. Jianguo Wu, Bingcheng Wang and Liu Fan contributedto the preparation and revision of the paper.

Acknowledgments

The authors are grateful to Dr. Qun Ma for her help with thepreparation of figures in this paper.

References

Agrawal, R., Imielinski, T., Swami, A.N., 1993. Mining association rules between setsof items in large databases. In: Proceedings of the 1993 ACM SIGMOD Inter-national Conference on Management of Data. pp. 207–216.

Agrawal, R., Srikant, R., 1994. Fast algorithms for mining association rules. In:Proceedings of 20th International Conference Very Large Data Bases.pp. 487–499.

Azapagic, A., 2004. Developing a framework for sustainable development indicatorsfor the mining and minerals industry. J. Clean. Prod. 12 (6), 639–662.

Bebbington, A.J., Bury, J.T., 2009. Institutional challenges for mining and sustain-ability in Peru. Proc. Natl. Acad. Sci. USA 106 (41), 17296–17301.

Chae, Y.M., Ho, S.H., Cho, K.W., Lee, D.H., Ji, S.H., 2001. Data mining approach topolicy analysis in a health insurance domain. Int. J. Med. Inform. 62 (2),103–111.

Corder, G.D., McLellan, B.C., Green, S.R., 2012. Delivering solutions for resourceconservation and recycling into project management systems through SUSOPs.Miner. Eng. 29, 47–57.

Giurco, D., Cooper, C., 2012. Mining and sustainability: asking the right questions.Miner. Eng. 29, 3–12.

Hajkowicz, S.A., Heyenga, S., Moffat, K., 2011. The relationship between mining andsocio-economic well being in Australia's regions. Resour. Policy 36 (1), 30–38.

Han, J., Kamber, M., Pei, J., 2011. Data Mining: Concepts and Techniques: Conceptsand Techniques. Elsevier.

Hastie, T., Tibshiran, R., Friedman, J.H., 2009. The Elements of Statistical Learning:Data Mining, Inference, and Prediction. Springer, New York.

Huang, L., Wu, J., Yan, L., 2015. Defining and measuring urban sustainability: a re-view of indicators. Landsc. Ecol. 30 (7), 1175–1193.

GRI, 2000. Sustainability reporting guidelines on economic, environmental, andsocial performance. Global Reporting Initiative.

GRI, 2012. Sustainability Reporting Guidelines Version 3.1. Global ReportingInitiative.

Kommadath, B., Sarkar, R., Rath, B., Fuzzy Logic, A., 2012. Based approach to assesssustainable development of the mining and minerals sector. Sustain. Dev. 20(6), 386–399.

Kotsiantis, S., Kanellopoulos, D., 2006. Association rules mining: a recent overview.GESTS Int. Trans. Comput. Sci. Eng. 32 (1), 71–82.

McLellan, B.C., Corder, G.D., Giurco, D., Green, S., 2009. Incorporating sustainabledevelopment in the design of mineral processing operations – review andanalysis of current approaches. J. Clean. Prod. 17 (16), 1414–1425.

McLellan, B.C., Corder, G.D., 2013. Risk reduction through early assessment andintegration of sustainability in design in the minerals industry. J. Clean. Prod.53, 37–46.

Mudd, G.M., 2007. The Sustainability of mining in Australia: Key Production Trendsand Their Evironmental Implications for the Future.

Phillips, J., 2012. Using a mathematical model to assess the sustainability of pro-posed bauxite mining in Andhra Pradesh, India from a quantitative-based en-vironmental impact assessment. Environ. Earth Sci. 67 (6), 1587–1603.

Solomon, F., Katz, E., Lovel, R., 2008. Social dimensions of mining: research, policyand practice challenges for the minerals industry in Australia. Resour. Policy 33(3), 142–149.

Sorensen, P., 2012. Sustainable development in mining companies in South Africa.Int. J. Environ. Stud. 69 (1), 21–40.

The State Council of the People's Republic of China, The nationwide resource-basedcities sustainable development planning (2013–2020). ⟨⟨http://www.gov.cn/zwgk/2013-12/03/content_2540070.htm⟩⟩, In 2013.

Tsay, Y.-J., Chiang, J.-Y., 2005. CBAR: an efficient method for mining associationrules. Knowl.-Based Syst. 18 (2), 99–105.

Tuazon, D., Corder, G., Powell, M., Ziemski, M., 2012. A practical and rigorous ap-proach for the integration of sustainability principles into the decision-makingprocesses at minerals processing operations. Miner. Eng. 29, 65–71.

Valdivia, S.M., Ugaya, C.M.L., 2011. Life cycle inventories of gold artisanal and small-scale mining activities in Peru. J. Ind. Ecol. 15 (6), 922–936.

Worrall, R., Neil, D., Brereton, D., Mulligan, D., 2009. Towards a sustainability cri-teria and indicators framework for legacy mine land. J. Clean. Prod. 17 (16),1426–1434.

Wu, J., 2010. Urban sustainability: an inevitable goal of landscape research. Landsc.Ecol. 25 (1), 1–4.

Yu, J., Yao, S., Chen, R., Zhu, K., Yu, L., 2005. A quantitative integrated evaluation ofsustainable development of mineral resources of a mining city: a case study ofHuangshi, Eastern China. Resour. Policy 30 (1), 7–19.

Yu, J., Zhang, Z., Zhou, Y., 2008. The sustainability of China's major mining cities.Resour. Policy 33 (1), 12–22.