Risk assessment for supply chain based on Cloud model

Abstract

In recent years, supply chain risk management has been followed with interest due to the short life cycle of products. How to identify risk indicators can help evaluate risks on supply chains. Commonly adopted methods such as Fuzzy to determine the level of risks have limitations. In this paper, a framework of supply chain risk evaluation is first proposed and risk indicators are identified by theoretical surveys from 35 keywords and empirical analysis from 448 questionnaires. Moreover, both linguistic risk assessment model and Cloud model are used to evaluate risks of supply chain. The Cloud model evaluation results are between general risk and high risk but closer to high risk. In addition, Cloud expected value of risk is 6.54 which is within the high-risk range, and evaluation results are also high risk. It is shown that when the weights are the same, the cloud model can determine the priority of risk indicators, and reflect volatility and randomness comparing with other evaluation methods.

Keywords

Cloud model supply chain risk management word frequency risk identification risk evaluation

1 Introduction

As a new type of management, supply chain achieves the unification of logistics, capital flow and information flow by coordinating its upstream and downstream enterprises [1]. Production and sales are not only the rules of the entire supply chain but also represent the integration of supply chain [2, 3]. Moreover, globalization and short life cycles of products lead to a more complex supply chain.. Traditional approaches to supplying chain management have undergone tremendous changes [4]. Although this shift of supply chain has enabled companies to compete effectively in their business 7 environment [5, 6], it has made supply chains more fragile [7]. Facing increasingly fierce competition environment, it is necessary to deal with the risks that may arise in the pursuit of efficiency in supply chain.

Based on the above facts, how to identify and evaluate risks will help to build a flexible supply chain network [8]. However, most of existing evaluation methods are using AHP-Fuzzy model and the ambiguity of risk factors is not well reflected. In actual evaluation, both selection of evaluation domain to calculate weight and generation of comprehensive evaluation matrix are randomness and uncertainty. This study tries to use Cloud model to substitute AHP-Fuzzy membership function to evaluate the risk in supply chains, especially the fuzzy and random of risk assessment model.

The purpose of this study is to establish a supply chain risk indicator system according to supply chain management theory and supply chain risk management theory, and specially, and to propose a Cloud model for supply chain risk evaluation. In the process of risk assessment, in order to fully consider the randomness and uncertainty of the evaluation, the fuzzy and random relationship between the risk classification of supply chain and its influencing factors is represented by Cloud model digital language, so as to eliminate the deviation caused by human factors and provide reasonable decision-making basis for supply chain risk assessment.

The structure of this paper is as follows. Section 2 focuses on literature review. Section 3 highlights framework development and research design for concerned studies. Research methodology is discussed in Section 4. Results and outcomes are described in Section 5. Discussions and findings are presented in Section 6. The final section presents conclusions and further research.

2 Literature review

2.1 Supply chain management

In the context of production globalization and knowledge economy, market competition has become increasingly fierce [9]. The increasing of production levels has promoted consumer purchasing and demand diversification [10]. In order to gain consumer recognition in new market environment, enterprises began to change their management thinking from pursuing interests of individual companies to form new types of cooperative relationship [11]. Thus, collaborative relationship between the upstream and downstream are given fully consideration in supply chain management.

In this circumstance, supply chain management faces new problems and risks. Christopher and Lee [12] pointed out that supply chain management is facing four serious challenges: new competition rules, industrial globalization, downward price pressure and strengthen customer control. The challenges along with the complex internal and external environment have caused difficulties in supply chain management. How to effectively identify and evaluate supply chain risks using various technologies is a topic facing both academic and business communities.

2.2 Supply chain risk management

At present, more and more scholars and managers have paid enough attention to supply chain risk management. However, an approbatory definition of supply chain risk management has not been formed. Christopher and Towill [13] define supply chain risk management as supply chain members jointly manage external risks through a collaborative mechanism to reduce the vulnerability of the whole supply chain. Juttner et al. [14] believes that identifying potential risk factors and taking corrective measures to control vulnerability is the purpose of supply chain risk management. Sodhi et al. [15] defines supply chain risk management as the coordination and collaboration risk management of supply chain members to ensure the overall benefits and continuity of supply chain. Although there is no consensus definition of supply chain risk management, there are many relatively research results on supply chain risk management.

Supply chain risk is unplanned and unexpected abnormalities that affect the coordination and collaboration of supply chain [16], and may even cause supply chain disruptions [17]. Supply chain risk management is a collaborative process to identify potential risk factors and mitigate the negative impact of risk through various technologies and methods [18]. At the same time, a systematic theoretical framework plays a vital role in guiding supply chain risk management. Analyzing and studying the relationship between these steps is of great significance for understanding the connotation of supply chain risk management.

2.3 Supply chain risk identification

Supply chain risk management focuses on risk identification, risk assessment and risk control [19]. Risk identification is the first step in supply chain risk management, and it refers to systematically understanding various risks and analyzing potential causes of risk events. Risk identification can help to establish an effective indicator system so that risk assessment can proceed smoothly. Table 1 shows summary of supply chain risk indicators by different authors.

Table 1
Summary of supply chain risk indicators

Author (s) Risk indicators

Diabat et al. [20] Product risk, information risk, management risk, macro risk, demand risk, supply risk

Cagliano et al. [21] Internal risk (strategic, tactical, operational), external risk (catastrophic, economic, social, political, legal, cultural, industrial, partner)

Avinash et al. [22] Supply risk, demand risk, operate risk, environmental risk

Lavastre et al. [23] Delivery risk, quality risk, non-compliance risk, supply disruption risk, production line disruption risk, price risk, costs risk, technical risk, forecasting risk, reliability risk, uncontrollable external risk, inventory risk, administrative risk, contract risk, customs risk, information system risk, commercial risk

Venkatesh et al. [24] globalization risk, product quality and raw material standards risk, lack of resources risk, supplier uncertainty risk, employees’ behaviour risk, infrastructure risk, delay risk, demand risk, customer displeasure risk, financial risk, security risk

Rogers et al. [25] cultural risk, operational risk, infrastructure risk, economic risk, forecasting risk, supplier related risk

Su et al. [26] Sustainable plan risk, communities for sustainability risk, sustainable operational process controlling risk, sustainable certificates and growth risk

Kilubi and Haasis [27] Supply risk, demand risk, inner and outer environmental risk

Song et al. [28] Demand and supply uncertainty risk, suppliers risk, responsiveness performance risk, inflexibility of supply source risk, quality risk, coordination risk, information sharing risk, sustainable technology risk, price and cost risk, exchange rates risk, market risk, natural disasters risk, environmental risk, human rights risk, social commitment risk, business ethics risk

Jiang et al. [29] Planning process risk, port service process risk, distribution process risk, relationship process risk, nuclear enterprise financial risk, and external environment-associated risk.

Prakash et al. [30] Supply risk, process risk, financial risk, demand risk

Xu et al. [31] Technical risk, market risk, environmental risk, social risk, operational risk

Author (s)	Risk indicators
Diabat et al. [20]	Product risk, information risk, management risk, macro risk, demand risk, supply risk
Cagliano et al. [21]	Internal risk (strategic, tactical, operational), external risk (catastrophic, economic, social, political, legal, cultural, industrial, partner)
Avinash et al. [22]	Supply risk, demand risk, operate risk, environmental risk
Lavastre et al. [23]	Delivery risk, quality risk, non-compliance risk, supply disruption risk, production line disruption risk, price risk, costs risk, technical risk, forecasting risk, reliability risk, uncontrollable external risk, inventory risk, administrative risk, contract risk, customs risk, information system risk, commercial risk
Venkatesh et al. [24]	globalization risk, product quality and raw material standards risk, lack of resources risk, supplier uncertainty risk, employees’ behaviour risk, infrastructure risk, delay risk, demand risk, customer displeasure risk, financial risk, security risk
Rogers et al. [25]	cultural risk, operational risk, infrastructure risk, economic risk, forecasting risk, supplier related risk
Su et al. [26]	Sustainable plan risk, communities for sustainability risk, sustainable operational process controlling risk, sustainable certificates and growth risk
Kilubi and Haasis [27]	Supply risk, demand risk, inner and outer environmental risk
Song et al. [28]	Demand and supply uncertainty risk, suppliers risk, responsiveness performance risk, inflexibility of supply source risk, quality risk, coordination risk, information sharing risk, sustainable technology risk, price and cost risk, exchange rates risk, market risk, natural disasters risk, environmental risk, human rights risk, social commitment risk, business ethics risk
Jiang et al. [29]	Planning process risk, port service process risk, distribution process risk, relationship process risk, nuclear enterprise financial risk, and external environment-associated risk.
Prakash et al. [30]	Supply risk, process risk, financial risk, demand risk
Xu et al. [31]	Technical risk, market risk, environmental risk, social risk, operational risk

2.4 Supply chain risk assessment

Risk assessment is the most important component of supply chain risk management. Supply chain risk assessment not only evaluates the risk issues within the core of supply chain but also evaluates the risks arising from other links in supply chain network. At present, a commonly used risk assessment method for scholars is to make a grade judgment of supply chain total risk level.

Previous studies have focused on methodologies to identify risk indicators and evaluate supply chain risks. Some of these studies on supply chain risk assessment are summarized in Table 2.

Table 2
Summary of supply chain risk assessment model

Methods Authors

Fuzzy-AHP Dong and Cooper [32], Aqlan [33]

Fuzzy-TOPSIS Mavia et al. [34], Junior et al. [35]

SEM Jennifer and Cross [36], Jain and Raj [37]

DEMATEL Cui et al. [38], Lin et al. [39], Song et al. [40]

FMEA Lo and Liou [41], Liu et al. [42], Fahimniaa et al. [43]

GA Rapeti et al. [44], Wang et al. [45], Hashemi et al. [46]

ANN Foo et al. [47], Behzadi et al. [48]

Methods	Authors
Fuzzy-AHP	Dong and Cooper [32], Aqlan [33]
Fuzzy-TOPSIS	Mavia et al. [34], Junior et al. [35]
SEM	Jennifer and Cross [36], Jain and Raj [37]
DEMATEL	Cui et al. [38], Lin et al. [39], Song et al. [40]
FMEA	Lo and Liou [41], Liu et al. [42], Fahimniaa et al. [43]
GA	Rapeti et al. [44], Wang et al. [45], Hashemi et al. [46]
ANN	Foo et al. [47], Behzadi et al. [48]

2.5 Literature summary

As noted, a research gap exists in supply chain risk evaluation models.

When FEMA method is used to evaluate supply chain risks, it first needs to set a certain score range for each risk indicator, and then calculate total score. This method is relatively easy to use but its evaluation is not strong for in fact risk indicators have a certain degree of each other.

The determination of weight is highly subjective when TOPSIS is used to evaluate supply chain risks, and the “best point” and the “worst point” are not easy to determine, so the distance of evaluation goals between “the best advantage” and “the worst point” cannot be determined.

The uncertainty and ambiguity of subjective cognition of experts by AHP-Fuzzy comprehensive evaluation are not convincing. In addition, there are problems such as large data statistics, difficult weight determination and complicated solution of eigenvectors when a large amount of risk indicators is existed in AHP-Fuzzy.

ANN evaluation results are single, and the number of network layers, the number of neuron nodes in each layer, the transfer function and selection of training algorithm are all based on a large number of experimental calculations without feasibility theoretical guidance.

The concept of “curve similarity” in GA (Grey Relational Analysis) theory is not clear. Some calculation formulas lack science and cannot meet the order-preserving effect when the original data is dimensionless.

The demand for SEM samples is large and calculation process is complicated. In actual application, there are existed selected principal components which are not representative of original variables, and the evaluation would be biased. Selected risk indicators are limited when DEMATE is used to evaluate supply chain risks. In fact, the number of supply chain risk indicators is very large and whether the selected risk indicators are representative is an issue to be considered. In addition, when scoring is based solely on subjective assumptions of experts, the obtained scoring results may not be objective and accurate.

Cloud model can simulate human mind to flexibly divide attribute space and complete conversion from quantitative values to qualitative concepts [49]. Also there is an overlapping between adjacent attribute values or languages, which makes the discovered knowledge robust [50]. Moreover, Cloud model can handle data with randomness and uncertainty well, so Cloud model can be introduced into supply chain risk evaluation [51]. Cloud model not only can reduce the impact of subjective indicators on supply chain risk evaluation results, but also realize uncertainty conversion between qualitative concepts and quantitative values to improve accuracy of evaluation results.

On the basis of previous research work, this paper applies cloud model theory to supply chain risk assessment. Based on the existing research results and literature, this paper selects representative and quantifiable factors that affect supply chain risk. Aiming at the problems of high redundancy and complex evaluation process of traditional supply chain risk evaluation indicators, the evaluation index set is reduced by theoretical exploration and empirical screening, and redundant indicators are eliminated to simplify the risk evaluation process.

3 Research framework and research design

In this section, research framework and research design put forward in this research were described.

3.1 Research framework

The detailed research framework of this study is shown in Fig. 1.

Fig. 1

Framework of study.

3.2 Research design

3.2.1 Theoretical exploring initial supply chain risk indicators

(1) Data collection

Theoretical exploring involves database selection, keyword selection and search, title and abstract review, and applying the inclusion and exclusion criteria. A single database is insufficient to retrieve all the references needed for supply chain risk indicators and this might lead to selection bias. In this study, Science Direct (SD), ProQuest (PQ) and Google Scholar (GS) databases are used to enhance the theoretical exploring. SD database has high-quality articles that are published in peer-reviewed journals indexed in SCIE, E-SCI and SSCI. PQ database is included as it is a multidisciplinary database with over 160 subject areas. To ensure a robust analysis without publication bias, GS is used to retrieve the grey literature. The search strings is “supply chain risk indicators” and Table 3 is the inclusion and exclusion criteria for selected articles.

Table 3
Inclusion and exclusion criteria for 258 selected articles

SD and PQ GS

Include scientific journals published only in English Include all published only in English

Select article that contains at least one keyword in each level of search string in the title, abstract or keyword Select article that contains at least one keyword in each level of search string in the entire article

Exclude book chapters, proceedings papers, periodicals and working papers Supply chain risk or relevant term should appear in the title, abstract or keyword

Exclude articles that are not relevant to supply chain risk evaluation Exclude articles’ citations lower than 10 criteria

Include period from 2010 to 2021 Exclude articles not relevant to supply chain risk evaluation

SD and PQ	GS
Include scientific journals published only in English	Include all published only in English
Select article that contains at least one keyword in each level of search string in the title, abstract or keyword	Select article that contains at least one keyword in each level of search string in the entire article
Exclude book chapters, proceedings papers, periodicals and working papers	Supply chain risk or relevant term should appear in the title, abstract or keyword
Exclude articles that are not relevant to supply chain risk evaluation	Exclude articles’ citations lower than 10 criteria
Include period from 2010 to 2021	Exclude articles not relevant to supply chain risk evaluation

Figure 2 illustrates the flowchart of theoretical exploring followed in this study.

Fig. 2

Flowchart of theoretical exploring.

To obtain the research focus of each cluster and to label them, the contents such as title, abstract, author keywords in each article are assessed using VOSViewer. VOSViewer is a tool for providing text mining capabilities for building and visualizing co-occurrence networks of important terms extracted from scientific literature. Figure 3 shows the keyword co-occurrence network structure.

Fig. 3

Keyword co-occurrence analysis of central cluster “supply chain risk indicators”.

(2) Word frequency statistics

Word frequency statistics is a literature measurement method that takes frequency of words as research object. It uses frequency statistical analysis of keywords to reveal core content and reflect evolution of research field. The content of supply chain risk indicators is broad and has close relations with sociology, economics and other disciplines. But some of them were not highly correlated with supply chain risk indicators, and some are overlapped with each other. We standardized keywords through manual screening and removed keywords that were not very professional and too broad meanings such as “Management”, “Indicator”, “Risk”, etc. Then synonymous and similar keywords were merged such as “Supply risk” and “Management risk” into “Supply management risk”, “Supply risk indicators” and “Risk indicators” into” Supply risk indicators “.

A total of 236 articles were retrieved and 35 keywords with the highest frequency of supply chain risk indicators were obtained, which is shown in Table 4.

Table 4

Word frequency statistics of risk indicators (F means Frequency)

Keywords	F	Keywords	F	Keywords	F
Political risk	36	Risk awareness of managers	5	cultural risk	3
Capital chain break	25	Net profit margin	5	forecasting risk	3
Partner trust risk	22	Force majeure	5	information sharing risk	3
Economic cycle	14	Customer preference change	4	responsiveness performance	2
Platform maintenance	11	Customer satisfaction	4	human rights risk	2
Information asymmetry	10	Vehicle failure	4	social commitment risk	2
Payment system stability	8	Safety stock level	4	exchange rates risk	2
Corporate social image	8	Capital turnover rate	4	business ethics risk	2
Defence against viruses	8	Supply chain flexibility	4	globalization risk	1
Total asset growth rate	8	Timeliness of delivery	4	infrastructure risk	1
Supply chain decision mistakes	6	Cargo damage rate	3	employees’ behaviour	1
Information transmission rate	5	Service reliability	3

(3) Brainstorming

Practical experience shows that brainstorming can eliminate compromise solution and find a group of feasible solutions through objective and continuous analysis of the discussed problems. Therefore, brainstorming has been widely used in decision-making. The results of brainstorming should be considered as the collective creation and the overall effect of all experts.

The brainstorming rules should be obeyed as follows:

Criticism and comments are prohibited, and ideas put forward by others must not be criticized or blocked;

The participants are fully relaxed and motivated by the assumptions of others, and fully focused on developing their own ideas;

Free speech is advocated, and the goal of the meeting is to seek for an increasing number of pursuit ideas;

Each participant should inspire or supplement himself with the ideas of others, or combine several ideas of others to come up with new ideas. All participants were equal and all ideas were written down.

Our brainstorming sessions are based on the following facts: invited members are new to this type of meeting; any other type of meeting might prevent some members from participating; members think on their own and the final group discussion will take place. At the beginning of brainstorming, the moderator briefly introduces the issues to be discussed, especially the rules of brainstorming. The brainstorming session lasted two and a half hours, with half an hour break in between so that participants could exchange ideas.

The expertise of brainstorming expert should be consistent with the decision making issues involved and include experts in the field of expertise, senior experts in other areas of expertise and experts with high logical thinking skills. Therefore, the result of the brainstorming is the collective creation of expert members and overall effect of the group’s mutual infection. In this paper, we design brainstorming in different fields, different disciplines and different levels of experts. So we invited university professors, graduate students, undergraduates and government officials to attend the meeting. A total of 11 people attended the seminar, including one moderator. Professional fields range from e-commerce, risk management, economics to ecological management research and other subjects.

In the process of brainstorming, there are usually many experts with different backgrounds proposing ideas, and the moderator is responsible for collecting the similar ideas to form corresponding groups. First, experts with different backgrounds put forward their own ideas. Then, several groups were formed. Each group selected a core expert as the team leader to fully discuss the proposed ideas. Finally, following the rules of brainstorming, new ideas by discussing previous ideas in or between groups are generated. After several rounds of discussion, a perfect solution finally emerged.

The one-factor variance analysis and T-test were performed on the experts’ variables of the gender, education background and degree participated in brainstorming, and the results showed that there were no significant differences in gender, education background and degree of the tested experts, as is shown in Table 5. The statistical analysis also verified that the results of brainstorming are credible.

Table 5

Respondents characteristics of 11 brainstorming experts

Characteristics	Group	SD	t	d	p
Gender	Male	8.27	1.590	0.188	0.115
	Female	8.25	1.118	0.219	0.267
Education back	Liberal	10.36	1.050	0.214	0.210
	Science	9.59	1.278	0.145	0.186
Degree	Undergraduate	12.92	1.064	0.262	0.290
	Postgraduate	13.65	1.719	0.259	0.110

Keywords with a frequency of less than 3 were selected and made judgments using brainstorming and a set of 10 important supply chain risk indicators was finalized, which was presented in Table 6.

Table 6

Brainstorming results of risk indicators

Indicators	EX1	EX2	EX3	EX4	EX5	EX6	EX7	EX8	EX9	EX10	EX11
Responsiveness performance	D	D	D	S	D	D	D	D	D	D	S
Human rights risk	S	D	D	D	S	D	D	S	D	D	D
Social commitment risk	D	D	D	S	D	D	D	D	D	S	D
Exchange rates risk	D	D	D	S	D	D	D	D	D	D	S
Business ethics risk	D	S	D	D	D	D	S	D	D	D	D
Globalization risk	D	D	D	D	D	D	D	D	S	D	D
Infrastructure risk	D	D	D	D	S	D	D	D	D	S	D
Employees’ behaviour	D	D	D	S	D	D	D	D	D	S	D

Note: Please provide your opinion based on given scale S: (save) or D: (delete).

Finally, risk indicators conducted by brainstorming were added to word frequency statistics and initial supply chain risk indicators were constructed (Table 7).

Table 7

Initial risk indicators of supply chain by theoretical exploring

Political risk	Capital chain break	Partner trust risk
Economic cycle	Platform maintenance	Information asymmetry
Payment system stability	Corporate social image	Defence against viruses
Total asset growth rate	Supply chain decision mistakes	Information transmission rate
Risk awareness of managers	Net profit margin	Force majeure
Logistics risk	Government incentives	Laws and regulations
Customer preference change	Customer satisfaction	Vehicle failure
Safety stock level	Capital turnover rate	Supply chain flexibility
Timeliness of delivery	Cargo damage rate	Service provider reliability

3.2.2 Empirical screening supply chain risk indicators

(1) Measurement grouping variables and instruments

Independent grouping variables are respondents’ genders, age, employment status, education, hierarchical level, expertise, treated for risk management, risk awareness.

The professional and technical expertise of respondent is an important variable because decisions will be formed in relative knowledge and expertise of respondent. The education level influences supply chain risk management in terms of uncertainty related to risk indicators identification and risk control. An efficient treatment for risk management and risk awareness can effectively reduce supply chain risk and it is a very important variable. The hierarchical level can modify the risk indicators and thus the risk’s management. The respondent’s age and employment status also influence their judgement towards supply chain risk.

The characteristics of respondents can affect risk characteristics (risk identification and number of risk indicators) and risk management (risk evaluation, risk rank and risk prevention) of supply chain.

(2) Statistical analysis of data

The first step was a confirmatory study with 500 different questionnaires which were organized between December 2019 and March 2020. The questionnaires were conducted by anonymous questionnaire and were electronically forwarded to respondents. The questionnaire is composed of three parts: characteristics of respondents (genders, age, employment status, education, hierarchical level, expertise, treated for risk management, risk awareness), risk indicators constructed by initial theoretical exploratory, general questions about respondent institution. The major factors in the questionnaires are risk indicators of supply chain by theoretical exploring (Table 7).

Data collection and descriptive statistical analysis such as average values and deviations of respondents’ questionnaires were analysed by SPSS 23.0. The accuracy of respondents’ questionnaires is examined by different statistical analysis: Pearson correlation, multiple linear regression analysis and discriminate analysis.

(3) Sample Characteristics

In total, 467 questionnaires were collected and 448 questionnaires were effectively recovered. The recovery rate of the questionnaire was 93.20% and the effective response rate was 96.14%. The research sample is non-random and appropriate. Therefore, it can be estimated that questionnaires represent the characteristics of the entire respondents. The respondents’ characteristics of 448 questionnaires are shown in Table 8.

Table 8
Respondents characteristics of 448 questionnaires

Characteristics Group N = 448 %

Gender male 324 72.32

female 124 27.68

Age <20 47 9.63

20–29 89 19.86

30–39 125 27.91

40–49 155 34.59

>50 32 17.64

Employment status employed 366 81.69

unemployed 20 4.46

retired 62 13.85

Education illiterate 6 1.34

primary 31 6.91

secondary 68 15.18

college and above 343 76.57

Hierarchical level General manager 88 19.64

Department manager 157 35.04

Functional supervisor 97 21.65

Production Line leader 56 12.50

General staff 50 11.17

Expertise university scholars 148 33.03

supply chain manager 107 23.88

risk control manager 133 29.68

student 60 13.41

Treated for risk management Yes 394 87.94

No 31 6.92

Not familiar 23 5.14

Risk awareness Yes 407 90.84

No 28 6.25

Not familiar 13 2.91

Characteristics	Group	N = 448	%
Gender	male	324	72.32
	female	124	27.68
Age	<20	47	9.63
	20–29	89	19.86
	30–39	125	27.91
	40–49	155	34.59
	>50	32	17.64
Employment status	employed	366	81.69
	unemployed	20	4.46
	retired	62	13.85
Education	illiterate	6	1.34
	primary	31	6.91
	secondary	68	15.18
	college and above	343	76.57
Hierarchical level	General manager	88	19.64
	Department manager	157	35.04
	Functional supervisor	97	21.65
	Production Line leader	56	12.50
	General staff	50	11.17
Expertise	university scholars	148	33.03
	supply chain manager	107	23.88
	risk control manager	133	29.68
	student	60	13.41
Treated for risk management	Yes	394	87.94
	No	31	6.92
	Not familiar	23	5.14
Risk awareness	Yes	407	90.84
	No	28	6.25
	Not familiar	13	2.91

(4) Reliability and validity analysis

According to statistical results of questionnaire survey, a reliability test was performed using SPSS 22.0. The reliability of scale is usually expressed by α reliability coefficient. For a high-reliability questionnaire, the reliability coefficient is greater than 0.8. In this paper, the reliability coefficient is 0.9151, which meets the reliability test requirements and proves the stability of the questionnaire (Table 9).

Table 9

Reliability analysis

Cronbach’s alpha	Cronbach’s alpha based on standardized items	N of items
0.9151	0.9120	27

Measurement validity refers to it that whether measured characteristic is target to be measured. The questionnaire in this study as created after extensive reading of relevant literature and field research. In the process of data collection, survey subjects were strictly selected. The preliminary statistical inferences of data also indicate that the measurement is valid.

(5) Membership analysis

The threshold value of membership analysis was conducted based on actual experience of 11 brainstorming experts. After comprehensive expert opinions, the threshold value determined by membership analysis was 0.4. Indicators which membership degree is greater than 0.4 (including 0.4) were retained (Table 10).

Table 10

Membership analysis

Indicator	M	Indicator	M	Indicator	M
Political risk	0.3778	Total asset growth rate	0.6142	Customer preference change	0.2879
Capital chain break	0.4851	Decision mistakes	0.4372	Customer satisfaction	0.5821
Partner trust risk	0.6162	Information transmission rate	0.5223	Vehicle failure	0.4708
Economic cycle	0.4851	Risk awareness of managers	0.5259	Safety stock level	0.2587
Platform maintenance	0.5539	Net profit margin	0.3254	Capital turnover rate	0.2079
Information asymmetry	0.4182	Force majeure	0.4836	Supply chain flexibility	0.2260
Payment system stability	0.4063	Logistics risk	0.4599	Timeliness of delivery	0.2234
Corporate social image	0.4181	Government incentives	0.4257	Cargo damage rate	0.3407
Defense against viruses	0.4539	Laws and regulations	0.3874	Service provider reliability	0.3236

M = membership degree.

(6) Correlation analysis

Correlation coefficient is Pearson correlation coefficient. The correlation analysis of survey data showed that there were significant correlations among five pairs of indicators (Table 11). According to the result of correlation analysis, five indicators with high correlation coefficient were eliminated.

Table 11

Correlation analysis

High membership degree	Low membership degree	Correlation coefficient
Information asymmetry	Information transmission rate	0.6899
Platform maintenance	Defines against viruses	0.7048
Logistics risk	Vehicle failure	0.7892
Customer satisfaction	Corporate social image	0.7437
Risk awareness of managers	Decision mistakes	0.842

3.3 Determination of supply chain risk indicators

A total of 12 indicators were finalized risk indicators of supply chain, which is shown in Table 12.

Table 12
Final supply chain risk indicators

Risk indicator system Risk indicator Risk factor

Operation Capital chain break Financial risk (RF1)

Partner trust risk Cooperative risk (RF2)

Logistics risk Logistics risk (RF3)

Information asymmetry Information risk (RF4)

Total asset growth rate Resource risk (RF5)

Environment Force majeure Natural risk (RF6)

Government incentives Policy risk (RF7)

Economic cycle Market risk (RF8)

Customer satisfaction Sales risk (RF9)

Risk awareness of managers Risk awareness (RF10)

Technology Payment system stability Security risk (RF11)

Platform maintenance Maintenance risk (RF12)

Risk indicator system	Risk indicator	Risk factor
Operation	Capital chain break	Financial risk (RF1)
	Partner trust risk	Cooperative risk (RF2)
	Logistics risk	Logistics risk (RF3)
	Information asymmetry	Information risk (RF4)
	Total asset growth rate	Resource risk (RF5)
Environment	Force majeure	Natural risk (RF6)
	Government incentives	Policy risk (RF7)
	Economic cycle	Market risk (RF8)
	Customer satisfaction	Sales risk (RF9)
	Risk awareness of managers	Risk awareness (RF10)
Technology	Payment system stability	Security risk (RF11)
	Platform maintenance	Maintenance risk (RF12)

The supply chain risk indicator system is composed of three sub-system aspects: operation, environment and technology. Operation risk mainly includes financial risk, cooperative risk, logistics risk, information risk and resource risk. Financial risk determines the sustainability and robustness of supply chain, and supply chain with capital risk has a certain risk of disruption. The cooperative risk in supply chain is if there is deviation of each other’s values, it is easy to lead to the disintegration of supply chain. Logistics risk such as time delay and efficiency reduction will lose the trust of downstream enterprises in supply chain, which leads to hindered cooperation. Information risk in supply chain mainly refers to supply chain decision-making mistakes which may be caused by the inaccuracy of information and the failure to reach the receiver in time.

Environment risk mainly includes natural risk, policy risk, market risk, sales risk and risk awareness. Natural risk affects every node of supply chain, thus affecting the stability of the whole supply chain. Policy risk not only affects the production and operation of upstream and downstream enterprises of supply chain, but also affects the stability and sustainability of supply chain. Market risk is the risk brought by price, supply and demand change of supply chain products or services, and the risk brought by the sufficiency, stability and price change of raw materials, spare parts and other supplies. Sales risk refers to the impact of price reduction, inventory, seasonal change and consumption habit on supply chain products due to the change of supply chain competitive environment. Risk awareness refers to whether supply chain can respond quickly and what strategies can be adopted to reduce the potential risks.

Technology risk includes security risk and maintenance risk. Security risk can affect the source product of supply chain processing, transportation and later tracing information, as well as processing information, transportation information, storage information, sales feedback information and other information transmission timeliness. Supply chain network needs to achieve information sharing, the higher the degree of sharing, the more the decision-making direction of supply chain is in line with the market demand, this is the maintenance risk.

4 Methodology

4.1 Cloud model

Cloud is a transformation model that uses linguistic values to represent the uncertainty between qualitative concepts and quantitative concepts in order to achieve the ambiguity and randomness of things in natural world [52, 53]. Cloud not only explains things from random theory and fuzzy theory, but also reflects the correlation between ambiguity and randomness, which constitutes a mapping between quantitative and qualitative [54].

Suppose U is a qualitative concept that contains precise numerical values, C represents the qualitative concept of U. If x ∈ U and x is a random realization of quantitative value C, μ (x) ∈ [0, 1] has a stable tendency to randomness. If: $μ : U \to [0, 1], \forall x \in U x \to μ (x)$ (1)

Then, the distribution of x in quantitative field U is called a cloud, and each x is called a Cloud drop.

The digital characteristics of cloud are reflected the integrity concept of Cloud based on normal distribution function and normal membership function. Three digital features Ex (Expected value), En (Entropy) and He (Hyper entropy) are represented the concept of Cloud.

Ex: In the universe space, Cloud drop is the point that can best represent the qualitative concept, and its expectation is the centre value in the universe space.

En: En is jointly determined by randomness and ambiguity of qualitative concept and reflects the degree of dispersion of Cloud drop [54]. It also reflects the value range of cloud drops that can be accepted by qualitative concepts in the universe of space. The larger the value ranges of cloud drops, the more fuzzy the qualitative concepts, which also reflects the correlation between randomness and ambiguity.

He: He is the uncertainty measure of En. It reveals the cohesion, uncertainty and ambiguity of all points in linguistic value.

4.2 Cloud generator

4.2.1 Forward cloud generator

The evaluation interval [αβ] of the comment object is divided into n sub-intervals, where the i sub-interval is $[λ_{i}^{min} λ_{i}^{max}]$ , the n sub-interval is input and the cloud parameters (Ex_i, En_i, He_i) are output, wherein: ${Ex}_{i} = {\begin{matrix} λ_{i}^{min} i = 1 \\ \frac{λ_{i}^{min} + λ_{i}^{max}}{2} 1 < i < n \\ λ_{i}^{max} i = n \end{matrix}, {En}_{i} = \frac{λ_{i}^{min} + λ_{i}^{max}}{3}, {He}_{i} = η$ (2)

η is a constant that reflects the randomness of the evaluation value. The value of η should not be too large because He_i becomes larger, Ex_i needs to be adjusted according to the degree of ambiguity of the comment itself (Fig. 4).

Fig. 4

Forward cloud generator.

4.2.2 Reverse cloud generator

The determination of cloud parameters is based on inverse cloud generator model. The purpose of reverse cloud generator is to generate three characteristics Ex, En and He of Cloud model. Enter sample points X_i = (x_i1, x_i2, ⋯ , x_im), i = 1, 2, ⋯ , n and m characteristics (Ex₁, Ex₂, ⋯ , Ex_m ; En₁, En₂, ⋯ , En_m ; He₁, He₂, ⋯ , He_m) of cloud (C₁, C₂, ⋯ , C_m) are output (Fig. 5).

Fig. 5

Reverse cloud generator.

The process is as follows:

(1) Membership calculation

$μ_{i} = e^{\frac{2 ln (1 / 2)}{n} \times i}$ (3)

(2) Ex_i calculation

$\begin{matrix} {\bar{x}}_{1} = \frac{1}{n} \sum_{i = 1}^{n} (x_{1 i} (\frac{1}{2} + μ_{i})), \dots, \\ {\bar{x}}_{m} = \frac{1}{n} \sum_{i = 1}^{n} (x_{mi} (\frac{1}{2} + μ_{i})) \end{matrix}$ (4) $({Ex}_{1}, {Ex}_{2}, \dots, {Ex}_{m}) = ({\bar{x}}_{1}, {\bar{x}}_{2}, \dots, {\bar{x}}_{m})$ (5)

(3) En_i calculation

$\begin{matrix} {En}_{1} = \sqrt{\frac{1}{n - 1} \sum_{i = 1}^{n} {(x_{1 i} - {Ex}_{1})}^{2}}, \dots, \\ {En}_{m} = \sqrt{\frac{1}{n - 1} \sum_{i = 1}^{n} {(x_{mi} - {Ex}_{m})}^{2}} \end{matrix}$ (6)

(4) He_i calculation

$\begin{matrix} {En}_{1}^{'} = \sqrt{\frac{- (x_{1 i} - {Ex}_{1})}{2 ln μ_{i}}}, \dots, \\ {En}_{m}^{'} = \sqrt{\frac{- (x_{mi} - {Ex}_{m})}{2 ln μ_{i}}} \end{matrix}$ (7)

$\begin{matrix} {He}_{1} = \sqrt{\frac{1}{n - 1} \sum_{i = 1}^{n} ({En}_{1 i}^{'} - {\bar{En}}_{1}^{'})}, \dots, \\ {He}_{m} = \sqrt{\frac{1}{n - 1} \sum_{i = 1}^{n} ({En}_{mi}^{'} - {\bar{En}}_{m}^{'})} \end{matrix}$ (8)

4.3 Cloud model and rules

Cloud computing rules are shown in Table 13.

Table 13
Cloud computing rules

Computing Ex En He

+ Ex₁ + Ex₂ $\sqrt{{En}_{1}^{2} + {En}_{2}^{2}}$ $\sqrt{{He}_{1}^{2} + {He}_{2}^{2}}$

– Ex₁ - Ex₂ $\sqrt{{En}_{1}^{2} + {En}_{2}^{2}}$ $\sqrt{{He}_{1}^{2} + {He}_{2}^{2}}$

× Ex₁ × Ex₂ $| {Ex}_{1} \times {Ex}_{2} | \sqrt{{[\frac{{En}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{En}_{2}}{{Ex}_{2}}]}^{2}}$ $| {Ex}_{1} \times {Ex}_{2} | \sqrt{{[\frac{{He}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{He}_{2}}{{Ex}_{2}}]}^{2}}$

÷ Ex₁ ÷ Ex₂ $| \frac{{Ex}_{1}}{{Ex}_{2}} | \sqrt{{[\frac{{En}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{En}_{2}}{{Ex}_{2}}]}^{2}}$ $| \frac{{Ex}_{1}}{{Ex}_{2}} | \sqrt{{[\frac{{He}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{He}_{2}}{{Ex}_{2}}]}^{2}}$

Computing	Ex	En	He
+	Ex₁ + Ex₂	$\sqrt{{En}_{1}^{2} + {En}_{2}^{2}}$	$\sqrt{{He}_{1}^{2} + {He}_{2}^{2}}$
–	Ex₁ - Ex₂	$\sqrt{{En}_{1}^{2} + {En}_{2}^{2}}$	$\sqrt{{He}_{1}^{2} + {He}_{2}^{2}}$
×	Ex₁ × Ex₂	$\| {Ex}_{1} \times {Ex}_{2} \| \sqrt{{[\frac{{En}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{En}_{2}}{{Ex}_{2}}]}^{2}}$	$\| {Ex}_{1} \times {Ex}_{2} \| \sqrt{{[\frac{{He}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{He}_{2}}{{Ex}_{2}}]}^{2}}$
÷	Ex₁ ÷ Ex₂	$\| \frac{{Ex}_{1}}{{Ex}_{2}} \| \sqrt{{[\frac{{En}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{En}_{2}}{{Ex}_{2}}]}^{2}}$	$\| \frac{{Ex}_{1}}{{Ex}_{2}} \| \sqrt{{[\frac{{He}_{1}}{{Ex}_{1}}]}^{2} + {[\frac{{He}_{2}}{{Ex}_{2}}]}^{2}}$

4.4 Cloud evaluation process

Cloud evaluation of supply chain risk is described as follows:

Define risk factors and evaluation criteria, then establishing factor sets and review sets.

Determine weight sets and obtaining the relative weights of different risk factors.

Generate comprehensive evaluations in the form of digital features of Cloud models by reverse cloud generator and weight coefficient matrix.

In the fuzzy Cloud comprehensive evaluation method, Cloud model is used to calculate weight coefficient matrix and comprehensive evaluation matrix without traditional membership function. First, distribution of weights is statistically analyzed, and then, the parameters (Ex₁, Ex₂, ⋯ , Ex_m) are derived using the inverse Cloud generator model. Thus, the obtained weight coefficient matrix is: $A = [a_{1}, a_{2}, \dots, a_{q}] = {(\begin{matrix} \begin{matrix} {Ex}_{a 1} \\ {Ex}_{a 2} \end{matrix} & \begin{matrix} {En}_{a 1} \\ {En}_{a 2} \end{matrix} & \begin{matrix} {He}_{a 1} \\ {He}_{a 2} \end{matrix} \\ ⋮ & \dots & ⋮ \\ \begin{matrix} {Ex}_{aq} \end{matrix} & \begin{matrix} {En}_{aq} \end{matrix} & \begin{matrix} {He}_{aq} \end{matrix} \end{matrix})}^{T}$ (9)

For comprehensive evaluation matrix, the risk scores are assigned to corresponding risk scales, and then, the statistical samples are converted into parameters (Ex₁, Ex₂, ⋯ , Ex_m) using inverse cloud generator. The comprehensive evaluation matrix is: $R = [\begin{matrix} r_{1} \\ r_{2} \\ ⋮ \\ r_{q} \end{matrix}] = (\begin{matrix} \begin{matrix} {Ex}_{1} \\ {Ex}_{2} \end{matrix} & \begin{matrix} {En}_{1} \\ {En}_{2} \end{matrix} & \begin{matrix} {He}_{1} \\ {He}_{2} \end{matrix} \\ ⋮ & \dots & ⋮ \\ \begin{matrix} {Ex}_{q} \end{matrix} & \begin{matrix} {En}_{q} \end{matrix} & \begin{matrix} {He}_{q} \end{matrix} \end{matrix})$ (10)

(4) Comprehensive evaluation of Cloud models of risk factors

$\begin{matrix} \begin{matrix} B = A \circ R = {(\begin{matrix} \begin{matrix} {Ex}_{a 1} \\ {Ex}_{a 2} \end{matrix} & \begin{matrix} {En}_{a 1} \\ {En}_{a 2} \end{matrix} & \begin{matrix} {He}_{a 1} \\ {He}_{a 2} \end{matrix} \\ ⋮ & \dots & ⋮ \\ \begin{matrix} {Ex}_{aq} \end{matrix} & \begin{matrix} {En}_{aq} \end{matrix} & \begin{matrix} {He}_{aq} \end{matrix} \end{matrix})}^{T} \circ (\begin{matrix} \begin{matrix} {Ex}_{1} \\ {Ex}_{2} \end{matrix} & \begin{matrix} {En}_{1} \\ {En}_{2} \end{matrix} & \begin{matrix} {He}_{1} \\ {He}_{2} \end{matrix} \\ ⋮ & \dots & ⋮ \\ \begin{matrix} {Ex}_{q} \end{matrix} & \begin{matrix} {En}_{q} \end{matrix} & \begin{matrix} {He}_{q} \end{matrix} \end{matrix}) \\ = [\begin{matrix} {Ex}_{a 1} \times {Ex}_{1} + {Ex}_{a 2} \times {Ex}_{2} + \dots + {Ex}_{aq} \times {Ex}_{q} \\ \sqrt{{(| {Ex}_{a 1} \times {Ex}_{1} | \sqrt{{[\frac{{En}_{a 1}}{{Ex}_{a 1}}]}^{2} + {[\frac{{En}_{1}}{{Ex}_{1}}]}^{2}})}^{2} + {(| {Ex}_{a 2} \times {Ex}_{2} | \sqrt{{[\frac{{En}_{a 2}}{{Ex}_{a 2}}]}^{2} + {[\frac{{En}_{2}}{{Ex}_{2}}]}^{2}})}^{2} + \dots + {(| {Ex}_{an} \times {Ex}_{n} | \sqrt{{[\frac{{En}_{aq}}{{Ex}_{aq}}]}^{2} + {[\frac{{En}_{q}}{{Ex}_{q}}]}^{2}})}^{2}} \\ \sqrt{{(| {Ex}_{a 1} \times {Ex}_{1} | \sqrt{{[\frac{{He}_{a 1}}{{Ex}_{a 1}}]}^{2} + {[\frac{{He}_{1}}{{Ex}_{1}}]}^{2}})}^{2} + {(| {Ex}_{a 2} \times {Ex}_{2} | \sqrt{{[\frac{{He}_{a 2}}{{Ex}_{a 2}}]}^{2} + {[\frac{{He}_{2}}{{Ex}_{2}}]}^{2}})}^{2} + \dots + {(| {Ex}_{an} \times {Ex}_{n} | \sqrt{{[\frac{{He}_{aq}}{{Ex}_{aq}}]}^{2} + {[\frac{{He}_{q}}{{Ex}_{q}}]}^{2}})}^{2}} \end{matrix}] \\ = {[\begin{matrix} B_{1} \\ B_{2} \\ ⋮ \\ B_{N} \end{matrix}]}^{T} = {(\begin{matrix} \begin{matrix} {Ex}_{1} \\ {Ex}_{2} \end{matrix} & \begin{matrix} {En}_{1} \\ {En}_{2} \end{matrix} & \begin{matrix} {He}_{1} \\ {He}_{2} \end{matrix} \\ ⋮ & \dots & ⋮ \\ \begin{matrix} {Ex}_{n} \end{matrix} & \begin{matrix} {En}_{n} \end{matrix} & \begin{matrix} {He}_{n} \end{matrix} \end{matrix})}^{T} \end{matrix} \end{matrix}$ (11)

The fuzzy Cloud comprehensive evaluation results are obtained from the above formulas, and the hierarchical and stability characteristics are analyzed. Generally, the following priority relationship can be used for sorting:

Ex should be compared first, the bigger Ex is, the better the ranking;

If Ex are the same, the smaller En is, the better ranking;

If Ex and En are the same, the smaller He is, the better ranking.

5 Results and outcomes

5.1 Establishing linguistic and weight set of risk indicators

Nine linguistic rating scales is used to deal the uncertainty of human’s subjective judgments and the linguistic matrix of risk indicators is developed with the help of experts’ judgments [55]. Linguistic set of risk indicators are defined: extremely low = L0, fairly low = L1, very low = L2, low = L3, medium = L4, high = L5, very high = L6, fairly high = L7, extremely high = L8. The results of linguistic set are shown in Table 14.

Table 14
Linguistic set of risk indicators

Expert EX 1 EX 2 EX 3 EX 4 EX 5 EX 6 EX 7

RF1 L6 L2 L1 L6 L4 L2 L6

RF2 L4 L6 L4 L6 L4 L4 L5

RF3 L2 L4 L6 L3 L5 L3 L4

RF4 L6 L4 L6 L4 L5 L5 L4

RF5 L4 L6 L6 L3 L5 L5 L3

RF6 L7 L3 L3 L4 L3 L1 L4

RF7 L8 L2 L2 L8 L3 L7 L8

RF8 L1 L6 L7 L4 L4 L5 L6

RF9 L6 L4 L2 L3 L4 L4 L3

RF10 L6 L6 L2 L4 L4 L5 L4

RF11 L7 L6 L2 L5 L4 L1 L5

RF12 L7 L1 L1 L6 L2 L3 L4

Expert	EX 1	EX 2	EX 3	EX 4	EX 5	EX 6	EX 7
RF1	L6	L2	L1	L6	L4	L2	L6
RF2	L4	L6	L4	L6	L4	L4	L5
RF3	L2	L4	L6	L3	L5	L3	L4
RF4	L6	L4	L6	L4	L5	L5	L4
RF5	L4	L6	L6	L3	L5	L5	L3
RF6	L7	L3	L3	L4	L3	L1	L4
RF7	L8	L2	L2	L8	L3	L7	L8
RF8	L1	L6	L7	L4	L4	L5	L6
RF9	L6	L4	L2	L3	L4	L4	L3
RF10	L6	L6	L2	L4	L4	L5	L4
RF11	L7	L6	L2	L5	L4	L1	L5
RF12	L7	L1	L1	L6	L2	L3	L4

Weight comparison matrix is established using the five rating scales and the specific meanings and codes are defined: extremely unimportant = C0, unimportant = C1, medium = C2, important = C3, extremely important = C4.

Seven business personals who were familiar with the business management process and expert in risk assessment were invited to evaluate risk indictors using different linguistic set. The results of weight set are shown in Table 15.

Table 15

Weight set of risk indicators

Indicator	RF1	RF2	RF3	RF4	RF 5	RF 6	RF7	RF8	RF9	RF10	RF11	RF12
RF1	–
RF2	C4
RF3	C2	C2
RF4	C1	C4	C1
RF5	C2	C1	C1	C2
RF6	C3	C3	C2	C2	C3
RF7	C1	C2	C3	C3	C3	C2
RF8	C2	C2	C2	C2	C2	C1	C4
RF9	C3	C3	C3	C3	C2	C2	C2	C1
RF10	C2	C2	C2	C4	C1	C1	C4	C4	C1
RF11	C1	C2	C2	C2	C1	C2	C1	C2	C1	C4
RF12	C1	C3	C3	C1	C1	C1	C3	C1	C1	C3	C2	–

5.2 Determining weight and linguistic Cloud matrices of risk indicators

In this step, the linguistic set and weight set of risk indicators are calculated to acquire Cloud matrix, which is shown in Table 16.

Table 16
Weight cloud matrix and linguistic cloud matrix

Indicator Weight cloud matrix Linguistic cloud matrix

Ex En He Ex En He

RF1 0.1027 0.8192 0.0220 4.8571 0.9081 0.0374

RF2 0.0994 0.8038 0.0357 6.1143 0.8957 0.0382

RF3 0.1032 0.7980 0.0254 6.1571 0.8901 0.0308

RF4 0.0234 0.8100 0.0203 5.1714 0.9013 0.0083

RF5 0.0997 0.8089 0.0006 6.2143 0.9026 0.0328

RF6 0.1061 0.7758 0.0053 5.0429 0.8920 0.0106

RF7 0.0132 0.7908 0.0259 6.3286 0.9117 0.0040

RF8 0.0930 0.7856 0.0130 5.9714 0.9142 0.0517

RF9 0.0933 0.7948 0.0176 5.0571 0.9040 0.0232

RF10 0.1277 0.8111 0.0113 5.5429 0.9138 0.0134

RF11 0.0532 0.7850 0.0218 5.4143 0.8947 0.0201

RF12 0.0851 0.7956 0.0223 4.8857 0.8866 0.0306

Indicator	Weight cloud matrix	Linguistic cloud matrix
RF1	0.1027	0.8192	0.0220	4.8571	0.9081	0.0374
RF2	0.0994	0.8038	0.0357	6.1143	0.8957	0.0382
RF3	0.1032	0.7980	0.0254	6.1571	0.8901	0.0308
RF4	0.0234	0.8100	0.0203	5.1714	0.9013	0.0083
RF5	0.0997	0.8089	0.0006	6.2143	0.9026	0.0328
RF6	0.1061	0.7758	0.0053	5.0429	0.8920	0.0106
RF7	0.0132	0.7908	0.0259	6.3286	0.9117	0.0040
RF8	0.0930	0.7856	0.0130	5.9714	0.9142	0.0517
RF9	0.0933	0.7948	0.0176	5.0571	0.9040	0.0232
RF10	0.1277	0.8111	0.0113	5.5429	0.9138	0.0134
RF11	0.0532	0.7850	0.0218	5.4143	0.8947	0.0201
RF12	0.0851	0.7956	0.0223	4.8857	0.8866	0.0306

5.3 Cloud model comprehensive evaluation of risk factors

The weight Cloud matrix and linguistic Cloud matrix of risk factors are represented by matrices A and R. Therefore, the final evaluation of supply chain risk Cloud matrix result is as follows: $\begin{matrix} B = A \circ R & = {[\begin{matrix} 0.1027 & 0.8192 & 0.0220 \\ 0.0994 & 0.8038 & 0.0357 \\ 0.1032 & 0.7980 & 0.0254 \\ 0.0234 & 0.8100 & 0.0203 \\ 0.0997 & 0.8089 & 0.0006 \\ 0.1061 & 0.7758 & 0.0053 \\ 0.0132 & 0.7908 & 0.0259 \\ 0.0930 & 0.7856 & 0.0130 \\ 0.0933 & 0.7948 & 0.0176 \\ 0.1277 & 0.8111 & 0.0113 \\ 0.0532 & 0.7850 & 0.0218 \\ 0.0851 & 0.7956 & 0.0223 \end{matrix}]}^{T} \\ \circ [\begin{matrix} 4.8571 & 0.9081 & 0.0374 \\ 6.1143 & 0.8957 & 0.0382 \\ 6.1571 & 0.8901 & 0.0308 \\ 5.1714 & 0.9013 & 0.0083 \\ 6.2143 & 0.9026 & 0.0328 \\ 5.0429 & 0.8920 & 0.0106 \\ 6.3286 & 0.9117 & 0.0040 \\ 5.9714 & 0.9142 & 0.0517 \\ 5.0571 & 0.9040 & 0.0232 \\ 5.5429 & 0.9138 & 0.0134 \\ 5.4143 & 0.8947 & 0.0201 \\ 4.8857 & 0.8866 & 0.0306 \end{matrix}] \\ = [\begin{matrix} 6.5400 & 0.5321 & 0.0112 \end{matrix}] \end{matrix}$

5.4 Cloud model comprehensive evaluation results

The risk intervals are divided into five categories: extremely low risk, low risk, medium risk, high risk, and very high risk. The Cloud parameters for each risk interval are shown in Table 17.

Table 17
Risk interval and standard cloud parameter

Risk value Semantic value Standard cloud parameter

[0,2] Extremely lower risk (1,0.55,0.01)

[2,4] Lower risk (3,0.47,0.01)

[4,6] Mediate risk (5,1.03,0.01)

[6,8] Higher risk (7,0.76,0.01)

[8,10] Extremely higher risk (9,0.62,0.01)

Risk value	Semantic value	Standard cloud parameter
[0,2]	Extremely lower risk	(1,0.55,0.01)
[2,4]	Lower risk	(3,0.47,0.01)
[4,6]	Mediate risk	(5,1.03,0.01)
[6,8]	Higher risk	(7,0.76,0.01)
[8,10]	Extremely higher risk	(9,0.62,0.01)

Figure 6 shows Cloud evaluation results of supply chain risk compared with standard cloud parameters.

Fig. 6

Evaluation result of supply chain risk.

It is clear that Cloud model evaluation results is between general risk and high risk but closer to high risk, the evaluation results are more strongly affected by high-risk cloud model. Therefore, risk level of supply chain is high risk. In addition, Cloud expected value of risk is 6.5400 which is within the high-risk range of [6, 8], and evaluation results are also high risk.

6 Discussions

6.1 Evaluation result comparison

In order to verify the reliability of Cloud model for supply chain risk assessment, this paper evaluates the factors and weights selected in Cloud model with the fuzzy comprehensive evaluation method, using conventional fuzzy comprehensive method, the linguistic set and weight set are also used to calculate the data in Cloud model, and the final evaluation result is: $B = A \circ R = {[\begin{matrix} 0.1202 \\ 0.0811 \\ 0.1236 \\ 0.0562 \\ 0.0852 \\ 0.1032 \\ 0.0375 \\ 0.0806 \\ 0.0811 \\ 0.1032 \\ 0.0610 \\ 0.0671 \end{matrix}]}^{T} \circ [\begin{matrix} 4.1714 \\ 4.6286 \\ 4.3714 \\ 4.6429 \\ 4.9429 \\ 3.6286 \\ 4.2143 \\ 5.0857 \\ 3.8571 \\ 4.8286 \\ 4.9143 \\ 3.1857 \end{matrix}] = 4.3662$

According to the risk factor evaluation form, the supply chain has a medium risk. The risk weight of Cloud evaluation model compared with conventional fuzzy comprehensive evaluation is shown in Table 18.

Table 18
Risk weight and rank comparison

Indicator Cloud evaluation model Fuzzy comprehensive evaluation

Ex En He Rank Weight Rank

RF1 0.1027 0.8192 0.0220 4 0.1202 4

RF2 0.0994 0.8038 0.0357 6 0.0811 6

RF3 0.1032 0.7980 0.0254 3 0.1236 3

RF4 0.0234 0.8100 0.0203 11 0.0562 11

RF5 0.0997 0.8089 0.0006 5 0.0852 5

RF6 0.1061 0.7758 0.0053 2 0.1032 1

RF7 0.0132 0.7908 0.0259 12 0.0375 12

RF8 0.0930 0.7856 0.0130 8 0.0806 8

RF9 0.0933 0.7948 0.0176 7 0.0811 6

RF10 0.1277 0.8111 0.0113 1 0.1032 1

RF11 0.0532 0.7850 0.0218 10 0.0610 10

RF12 0.0851 0.7956 0.0223 9 0.0671 9

Indicator	Cloud evaluation model	Fuzzy comprehensive evaluation
RF1	0.1027	0.8192	0.0220	4	0.1202	4
RF2	0.0994	0.8038	0.0357	6	0.0811	6
RF3	0.1032	0.7980	0.0254	3	0.1236	3
RF4	0.0234	0.8100	0.0203	11	0.0562	11
RF5	0.0997	0.8089	0.0006	5	0.0852	5
RF6	0.1061	0.7758	0.0053	2	0.1032	1
RF7	0.0132	0.7908	0.0259	12	0.0375	12
RF8	0.0930	0.7856	0.0130	8	0.0806	8
RF9	0.0933	0.7948	0.0176	7	0.0811	6
RF10	0.1277	0.8111	0.0113	1	0.1032	1
RF11	0.0532	0.7850	0.0218	10	0.0610	10
RF12	0.0851	0.7956	0.0223	9	0.0671	9

The most important risk indicator is known to be risk awareness (RF10) and natural risk (RF6) which is found to be the least important one. Hence, risk awareness should be given the top priority in supply chain risk management. Moreover, Fuzzy comprehensive evaluation methods fail to effectively rank risk indicators in supply chains. Because risk evaluation is needed to be able to identify risk indicators, risk identification is the key step of supply chain risk management. Cloud model not only evaluates risk but also rank risks in supply chain.

The evaluation results are consistent with the Cloud model evaluation results. The results show that the model is feasible and effective, and the model considers the disturbance factors of supply chain risk indicators, which is closer to the actual situation.

6.2 Findings

Supply chain risk is a qualitative concept, which is difficult to be measured by conventional mathematical model. However, Cloud model can quantitatively express the qualitative classification of supply chain risk with Cloud digital features. The positive normal cloud generator of Cloud model can represent the randomness and fuzziness of supply chain risk classification by quantitative value of certainty calculation, and can better reflect the conversion relationship between qualitative language and quantitative value uncertainty in the process of supply chain risk level evaluation. Besides, its application is reflected on the risk evaluation in this paper.

(1) Cloud evaluation has better flexibility and ranking priority

Fuzzy comprehensive evaluation cannot determine the priority order of factor weights when weights are the same, so the similarities and differences between the two cannot be explained (Fig. 7). Therefore, in actual evaluation, there are ambiguity and randomness from the selection of evaluation domain in the calculation of weight coefficient and the generation of comprehensive evaluation matrix. Cloud model not only ranks the evaluation results but also analyzes the inherent uncertainty of evaluation objects, thus providing richer reference information for risk assessment.

Fig. 7

Rank of supply chain risk indicators.

(2) Cloud valuation has better randomness and robustness

Results obtained by conventional Fuzzy evaluation method are not reflected in the volatility and randomness. Moreover, in order to compare the simple results, the final evaluation result is simplified into a conversion score, which does not reflect the vague essence (Fig. 8).

Fig. 8

Fluctuation of supply chain risk indicators.

It can be seen from the evaluation results that Cloud evaluation model is consistent with evaluation results of existing methods, which shows that Cloud comprehensive evaluation method is reasonable and feasible. Moreover, Cloud fuzzy comprehensive evaluation method has volatility and randomness, reflecting the essence of ambiguity.

7 Conclusions

In this paper, the Cloud model theory is introduced, and the characteristics and advantages of the transformation between qualitative language and quantitative value are applied to the evaluation method of supply chain risk level. The Cloud model of supply chain risk level evaluation can not only realize the uncertain mapping between the value of supply chain risk factor and the qualitative description of supply chain risk level category, but also can convert supply chain qualitative risk level to corresponding Cloud model levels. This model has two advantages:

The model is to make full use of the uncertainty transformation mechanism of Cloud model, which makes the evaluation system take into account both subjectivity and objectivity, and has better flexibility compared with the traditional supply chain risk evaluation model.

The supply chain risk rating evaluation model based on cloud model realizes the transformation between the supply chain risk level category and the supply chain risk factor value, and to a certain extent, modifies the influence of subjective factors in the evaluation and the subjectivity of the evaluation results. In addition, it further explores the main influencing factors of the evaluation results, so as to make the comprehensive evaluation results objective and reliable, which is conducive to the follow-up optimization and avoidance of risks. Compared with Fuzzy mathematics, Cloud model can consider both fuzziness and randomness.

However, the algorithm of reverse Cloud generator is based on statistical principle, and approximately deduces the overall characteristics with the characteristics of samples. There are some errors in this process, which need to be studied and improved by later scholars. Besides, This paper gives a theoretical evaluation method of supply chain risk based on Cloud model. Although it has been verified by empirical research, the basis of index system construction is still not complete. It is expected that later scholars’ research can improve the basis of index system construction.

Footnotes

Acknowledgments

This work was financially supported by Social Science Foundation of Shandong Province (Grant No. 19BJCJ20).

References

Lafkihi

, Pan

and Ballot

, Freight transportation service procurement: a literature review and future research opportunities in omni channel E-commerce, Transportation Research Part E: Logistics and Transportation Review 125(5) (2019), 348–365.

Rofin

T.M.

and Mahanty

, Optimal dual-channel supply chain configuration for product categories with different customer preference of online channel, Electronic Commerce Research 18(3) (2018), 507–536.

Choshin

and Ghaffari

, An investigation of the impact of effective factors on the success of e-commerce in small- and medium-sized companies, Computers in Human Behaviour 66(1) (2017), 67–74.

Gregory

G.D.

, Ngo

L.V.

and Karavdic

, Developing e-commerce marketing capabilities and efficiencies for enhanced performance in business-to-business export ventures, Industrial Marketing Management 78(4) (2019), 146–157.

Yang

Z.F.

, Shi

and Yan

, Analysis on pure e-commerce congestion effect, productivity effect and profitability in China, Socio-Economic Planning Sciences 57(3) (2017), 35–49.

Kabanda

and Brown

, A structuration analysis of Small and Medium Enterprise (SME) adoption of E-Commerce: The case of Tanzania, Telematics and Informatics 34(7) (2017), 118–132.

Sreedevi

and Saranga

, Uncertainty and supply chain risk: The moderating role of supply chain flexibility in risk mitigation, International Journal of Production Economics 193(11) (2017), 332–342.

Margolis

J.H.

, Sullivan

K.M.

, Mason

S.J.

, et al., A multi-objective optimization model for designing resilient supply chain networks, International Journal of Production Economics 204(10) (2018), 174–185.

Park

, Bellamy

M.A.

and Basole

R.C.

, Structural anatomy and evolution of supply chain alliance networks: a multi-method approach, Journal of Operations Management 63(11) (2018), 79–96.

10.

Liu

and Wang

, Supply chain network equilibrium with strategic financial hedging using futures, European Journal of Operational Research 272(3) (2019), 962–978.

11.

Cortinhal

M.J.

, Lopes

M.J.

and Melo

M.T.

, A multi-stage supply chain network design problem with in-house production and partial product outsourcing, Applied Mathematical Modelling 70(6) (2019), 572–594.

12.

Christopher

and Lee

, Mitigating supply chain risk through improved confidence, International Journal of Physical Distribution & Logistics Management 34(5) (2004), 388–396.

13.

Christopher

and Towill

D.R.

, Developing market specific supply chain strategies, International Journal of Logistics Management 13(1) (2002), 1–14.

14.

Juttner

, Peck

and Christopher

, Supply chain risk management: outlining an agenda for future research, International Journal of Logistics: Research and Applications 6(4) (2003), 197–210.

15.

Sodhi

M.S.

, Son

B.G.

and Tang

C.S.

, Researchers’ perspectives on supply chain risk management, Production and Operations Management 21(1) (2012), 1–13.

16.

Aqlan

and Lam

S.S.

, A fuzzy-based integrated framework for supply chain risk assessment, International Journal of Production Economics 161(3) (2015), 54–63.

17.

Tang

and Musa

S.N.

, Identifying risk issues and research advancements in supply chain risk management, International Journal of Production Economics 133(1) (2011), 25–34.

18.

Aqlan

, A software application for rapid risk assessment in integrated supply chains, Expert Systems with Applications 43(1) (2016), 109–116.

19.

Clarke

and Boersma

, The governance of global value chains: unresolved human rights environmental and ethical dilemmas in the Apple supply chain, Journal of Business Ethics 143(1) (2017), 111–131.

20.

Diabat

, Govindan

and Panicker

V.V.

, Supply chain risk management and its mitigation in a food industry, International Journal of Production Research 50(11) (2012), 3039–3050.

21.

Cagliano

A.C.

, Marco

D.A.

, Grimaldi

and Rafele

, An integrated approach to supply chain risk analysis, Journal of Risk Research 15(7) (2012), 817–840.

22.

Avinash

, Viipul

and Felix

C.T.S.

, Quantifying risks in a supply chain through integration of fuzzy AHP and fuzzy TOPSIS, International Journal of Production Research 51(8) (2013), 2433–2442.

23.

Lavastre

, Gunasekaran

and Spalanzani

, Effect of firm characteristics, supplier relationships and techniques used on supply chain risk management (SCRM): an empirical investigation on French industrial firms, International Journal of Production Research 52(11) (2014), 3381–3403.

24.

Venkatesh

V.G.

, Rathi

and Patwa

, Analysis on supply chain risks in Indian apparel retail chains and proposal of risk prioritization model using Interpretive structural modelling,, Journal of Retailing and Consumer Services 26(9) (2015), 153–167.

25.

Rogers

, Srivastava

, Pawar

K.S.

and Shah

, Supply chain risk management in India e practical insights, International Journal of Logistics-Research and Applications 18(6) (2015), 1–22.

26.

C.M.

, Horng

D.J.

, Tseng

M.L.

, et al., Improving sustainable supply chain management using a novel hierarchical grey-DEMATEL approach, Journal of Cleaner Production 134(10) (2016), 469–481.

27.

Kilubi

and Haasis

H.D.

, Supply chain risk management research: avenues for further studies, International Journal of Supply Chain and Operations Resilience 2(1) (2016), 51–71.

28.

Song

, Ming

and Liu

H.C.

, Identifying critical risk factors of sustainable supply chain management: a rough strength-relation analysis method, Journal of Cleaner Production 143(2) (2017), 100–115.

29.

Jiang

, Li

and Shen

S.Y.

, Supply Chain Risk Assessment and Control of Port Enterprises: Qingdao port as case study, The Asian Journal of Shipping and Logistics 34(3) (2018), 198–208.

30.

Prakash

, Agarwal

and Kumar

, Risk assessment in automobile supply chain, Materials Today: Proceedings 5(2) (2018), 3571–3580.

31.

, Cui

Y.Y.

, Hu

, et al., Supply chain sustainability risk and assessment, Journal of Cleaner Production 225(7) (2019), 857–867.

32.

Dong

Q.X.

and Cooper

, An orders of magnitude AHP supply chain risk assessment framework, International Journal of Production Economics 182(12) (2016), 144–156.

33.

Aqlan

and Lam

S.S.

, Supply chain risk modelling and mitigation, International Journal of Production Research 53(18) (2015), 5640–5656.

34.

Mavia

R.K.

, Gohb

and Mavi

N.K.

, Supplier selection with Shannon entropy and Fuzzy-TOPSIS in the context of supply chain risk management, Procedia-Social and Behavioural Sciences 235(11) (2016), 216–225.

35.

Junior

F.R.

, Osiro

and Carpinetti

L.R.

, A comparison between Fuzzy-AHP and Fuzzy-TOPSIS methods to supplier selection, Applied Soft Computing 21(8) (2014), 194–209.

36.

Jennifer

and Cross

J.A.

, Systematic mechanism for identifying the relative impact of supply chain performance areas on the overall supply chain performance using SCOR model and SEM, International Journal of Production Economics 201(7) (2018), 102–115.

37.

Jain

and Raj

, Modelling and analysis of FMS performance variables by ISM, SEM and GTMA approach, International Journal of Production Economics 171(1) (2016), 84–96.

38.

Cui

, Chan

H.K.

, Zhou

Y.Z.

, et al., Exploring critical factors of green business failure based on Grey-Decision Making Trial and Evaluation Laboratory (DEMATEL), Journal of Business Research 98(5) (2019), 450–461.

39.

Lin

S.S.

, Li

C.B.

, Xu

F.Q.

, et al., Risk identification and analysis for new energy power system in China based on D numbers and decision-making trial and evaluation laboratory (DEMATEL), Journal of Cleaner Production 183(4) (2018), 81–96.

40.

Song

and Zhu

J.J.

, A multistage risk decision making method for normal cloud model considering behaviour characteristics, Applied Soft Computing 78(5) (2019), 393–406.

41.

H.W.

and Liou

J.H.

, A novel multiple criteria decision making based FMEA model for risk assessment, Applied Soft Computing 73(12) (2018), 684–696.

42.

Liu

H.C.

, You

J.X.

, Li

, et al., Failure mode and effect analysis under uncertainty: an integrated multiple criteria decision making approach, IEEE Transactions on Reliability 65(3) (2016), 1380–1392.

43.

Fahimniaa

, Christopher

, Tangb

, et al., Quantitative models for managing supply chain risks: a review, European Journal of Operational Research 247(1) (2015), 1–15.

44.

Rapeti

, Pasam

V.K.

, Gurram

K.R.

and Revuru

R.S.

, Performance evaluation of vegetable oil based nano cutting fluids in machining using grey relational analysis-A step towards sustainable manufacturing, Journal of Cleaner Production 172(1) (2018), 2862–2875.

45.

Wang

, Zhu

Z.Q.

and Wang

Y.H.

, A novel hybrid MCDM model combining the SAW, TOPSIS and GRA methods based on experimental design, Information Sciences 348(6) (2016), 27–45.

46.

Hashemi

S.H.

, Karimi

and Tavana

, An integrated green supplier selection approach with analytic network process and improved Grey relational analysis, International Journal of Production Economics 159(1) (2015), 178–191.

47.

Foo

P.Y.

, Lee

V.H.

, Tan

G.W.H.

and Ooi

K.B.

, A gateway to realising sustainability performance via green supply chain management practices: A PLS–ANN approach, Expert Systems with Applications 107(10) (2018), 1–14.

48.

Behzadi

, Sullivan

M.J.

, Olsen

T.L.

and Zhang

, Agribusiness supply chain risk management: A review of quantitative decision models, Omega 79(9) (2018), 21–42.

49.

Yan

and Xu

K.L.

, Methodology and case study of quantitative preliminary hazard analysis based on cloud model, Journal of Loss Prevention in the Process Industries 60(7) (2019), 116–124.

50.

Mardan

, Govindan

, Mina

, et al., An accelerated benders decomposition algorithm for a bi-objective green closed loop supply chain network design problem, Journal of Cleaner Production 235(10) (2019), 1499–1514.

51.

Z.M.

, Sun

X.K.

, Wang

Y.X.

, et al., Green supplier selection in straw biomass industry based on cloud model and possibility degree, Journal of Cleaner Production 209(2) (2019), 995–1005.

52.

, Wang

and Su

, Fuzzy-FMECA risk evaluation and its applications in Chinese train control systems based on cloud model, Journal of Intelligent and Fuzzy Systems 37(10) (2019), 1–11.

53.

Zhang

, Zhou

, Hu

, et al., Risk evaluation and analysis of a gas tank explosion based on a vapor cloud explosion model: A case study, Engineering Failure Analysis 101(7) (2019), 22–35.

54.

J.H.

, Chang

C.W.

Electronic Commerce Research 17(4) (2017), 627–660 Analyzing the operation of cloud supply cha, adotion barriers and business model.

55.

Ali

S.M.

, Moktadir

M.a.

, Kabir

, et al., Framework for evaluating risks in food supply chain: Implications in food wastage reduction, Journal of Cleaner Production 228(8) (2019), 786–800.

Risk assessment for supply chain based on Cloud model

Abstract

Keywords

1 Introduction

2 Literature review

2.1 Supply chain management

2.2 Supply chain risk management

2.3 Supply chain risk identification

3 Research framework and research design

3.1 Research framework

3.2.1 Theoretical exploring initial supply chain risk indicators

4.1 Cloud model

4.2.1 Forward cloud generator

5.1 Establishing linguistic and weight set of risk indicators

5.4 Cloud model comprehensive evaluation results

Table 17 Risk interval and standard cloud parameter Risk value Semantic value Standard cloud parameter [0,2] Extremely lower risk (1,0.55,0.01) [2,4] Lower risk (3,0.47,0.01) [4,6] Mediate risk (5,1.03,0.01) [6,8] Higher risk (7,0.76,0.01) [8,10] Extremely higher risk (9,0.62,0.01)

6.1 Evaluation result comparison

Footnotes

Acknowledgments

References

Table 17
Risk interval and standard cloud parameter

Risk value Semantic value Standard cloud parameter

[0,2] Extremely lower risk (1,0.55,0.01)

[2,4] Lower risk (3,0.47,0.01)

[4,6] Mediate risk (5,1.03,0.01)

[6,8] Higher risk (7,0.76,0.01)

[8,10] Extremely higher risk (9,0.62,0.01)