A novel belief-degree-based uncertain Tchebycheff norm DEA model for the case study of risks prioritizing in e-business projects Authors: Masoomeh Shahbazi, Hamidreza Taghizadeh Rajaei, Kazem Askarifar, Ali Mahmoodirad, Sadegh Niroomand Published: 2026 ABSTRACT AbstractE-business projects may face several risks that are potential to affect the success of such projects. As an important issue for a project manager in this field, the potential risks should be prioritized according to their potential impact on the success of the project. In this study, the potential risks of the e-business sector are to be evaluated and ranked based on the success criteria of such projects. For this aim a complete set of risks compared to the literature is considered and their impact on the project success criteria are determined linguistically by the experts of the field as a case study. To respect the uncertain nature of the evaluations, the linguistic evaluations are converted to belief-degree-based uncertain values. Then for the first time, a data envelopment analysis (DEA) model is used to evaluate and rank such risks. For this aim, the classical Tchebycheff norm DEA model is extended to a belief-degree-based uncertain environment for the first time. An extensive computational study including sensitivity analysis and comparative study is performed by the uncertain Tchebycheff norm model to evaluate and rank the risks of the case study. Based on the obtained results “introducing new technology” and “cultural risk” are the most and least important risks respectively.Keywords: E-business project, Risk evaluation and ranking, Uncertain data envelopment analysis, Tchebycheff norm model, Belief-degree-based uncertaintySubject terms: Mathematics and computing, Risk factors Introduction Conceptual background E-business, or even called electronic business, is a type of business to perform commercial activities and transactions over the internet. It consists of a wide range of online operations such as online retail, electronic payments, digital marketing, etc. Bai et al.1, Dogan2. E-business has transformed classical business methods by enabling companies to easily access global markets, decrease operational costs, and increase customer experiences3. In existence of digital platforms, businesses can simplify processes, increase efficiency, and provide personalized services. E-business also provides challenges like cybersecurity threats, data privacy problems, and the need for continuous technological conformity. Effective and good e-business strategies need robust infrastructure, secure payment methods and systems, and a strong online presence to make trust with customers4. As the digital economy is growing, e-business is becoming an important part of modern commerce, providing competitive and dynamic businesses to adapt to today’s changing market dynamics. By admitting digital transformation, companies can discover new opportunities, optimize and improve operations, and deliver more value to customers in today’s world. E-business project management includes the planning, execution, and control of projects which are developing or enhancing online business operations5. It needs a strategic method to align digital inventions with business goals, ensuring projects are completed on time, with the pre-determined budget, and to the preferred quality standards. Effective and acceptable e-business project management starts with a clear understanding of the project scope, goals, and stakeholder preferences. This involves determining key deliverables, like website development, e-commerce platform integration, or digital marketing teams. Project managers also should direct challenges to the digital environment such as quickly changing technologies, cybersecurity problems, and the need for continuous user experiences. Collaboration among cross-functional teams such as IT, marketing, and operations, is necessary to be sure that all aspects of the project are coordinated. Agile methodologies are usually used to adjust to evolving needs and market conditions. In addition, robust risk management practices are important to relieve potential disruptions such as technical failures or data violations. Effective e-business project management not only considers innovation and efficiency but also elevates customer satisfaction and competitive advantage of a project. E-business projects are naturally potential for a variety of risks that may significantly affect their success6,7. One of the initial risks is technical failure, which involves system crashes, software bugs, integration issues, etc. These can disturb and interrupt operations, cause downtime, and decrease revenue. Cybersecurity threats are also a risk as e-business projects usually work with sensitive customer data that can be attractive targets for hackers. Data violations can cause financial losses, legal penalties, and harm to the company’s reputation. Financial risks are also popular which are including budget overruns and unexpected costs, that may pressure resources and extend project timelines. Market risks such as changes in consumer preferences, extreme competition, and economic fluctuations, may decrease demand and profitability. Operational risks like inadequate resource allocation, poor team expertise, and ineffective processes, can negatively affect project execution and results in weak outcomes. In addition, supply chain interruptions that can be because of geopolitical issues, natural disasters, or supplier failures, may affect the availability of main resources. In order to manage these risks, e-business projects need suitable planning, correct risk management strategies, and continuous controlling and monitoring. Project success measures such as project time, project cost, project quality, etc. can be used to address challenges, evaluate risks, and ensure the successful implementation of e-business projects. By introducing potential risks, prioritizing them, and developing possible plans, project managers could improve resilience, optimize resource allocation, and improve the likelihood of achieving project goals. About the problem of risk evaluation and ranking, multi-criteria decision making (MCDM) and data envelopment analysis (DEA) approaches propose philosophically distinct and complement methodologies. MCDM methods like AHP8 or TOPSIS9, are typically subjective and rely on expert opinion to determine weight values for various risk factors and also produce a comprehensive rank of each entity according to their perceived weakness or threat degree. On the other hand, DEA is a non-parametric and data-driven method that determines the relative efficiency of decision-making units (shown as DMUs) by simultaneously considering multiple inputs and outputs for the DMUs. For the case of risk assessment, where risk factors behave as undesirable inputs or outputs, typical advanced DEA approaches provide some benefits. For example, super-efficiency DEA model10 allows to rank efficient DMUs or all DMUs or cross-efficiency DEA model11 enhances distinction by evaluating each DMU considering the optimal weights of other DMUs and promotes a peer-evaluation structure. In such area, network DEA model12 offers an internal viewpoint to assess the efficiency of sub-processes or stages of a system. The non-radial DEA models like the slack-based measure (SBM) mainly address input excesses and output shortness to provide a more accurate reflection of risk degrees in the cases that improvements are not proportional. To address uncertain nature of real-world data, advanced DEA models may have some critical methodological robustness. For example, fuzzy DEA13 considers possibility theory and allows for the incorporation of imprecise or vague data (e.g. linguistic risk assessments) by considering fuzzy values for inputs and outputs. On the other hand, stochastic DEA models14 apply probability theory to manage randomness of data. In cases with distributional ambiguity, various robust DEA models15 provide worst-case-oriented approaches. Literature review There are many studies in the literature focusing on evaluation and prioritizing the risks of e-business projects. Here, some of these studies are mentioned. Evangelidis et al.16 introduced risk assessment factors and success factors for e-government in a UK establishment. Kleist17, proposed an approach for evaluating e-business information systems projects. Stevens and Fowell6, performed a study to discover some perspectives on e-business software project risks. Currie18, proposed a knowledge-based risk assessment framework in order to evaluate web-enabled application outsourcing projects. Dimitrakos19, proposed the CORAS approach that integrates model-based security risk management into eBusiness systems development. Hunter et al.20 proposed a framework based on risk importance and probability for classification of business-to-business buying decisions for e-business benefits. Scott21, performed a study to measure the dimensions of perceived e-business risks. Lesjak and Vehovar22, introduced some factors that are important to evaluate e-business projects. Khokhar et al.23 performed an evidential reasoning based study for risk assessment of e-commerce projects. Sutton et al.24 performed a study on risk analysis in extended enterprise environments. Fritz and Canavari25, developed a management model for perceived e-business risks in food‐supply networks. They applied e‐trust as prerequisite for supply‐chain system innovation. In this study they identified some critical risk factors in B2B e-commerce relationships. Kefallinos et al.26 proposed an extended risk assessment model for secure E-government projects. Hussain et al.27 introduced a transactional risk-based decision-making system in e-business interactions. Zirakja and Samizadeh28, applied a fuzzy logic based approach for risk analysis in e-commerce. Hussain et al.29 proposed a probability based assessment method for financial risks in e-business associations. Wang et al.30 introduced a method using big data analytics concept to evaluate enterprise credit risk of e-Business platform. Eltweri et al.31 performed a study and discovered the important role of risk management in auditing e-business. Further than the above-mentioned studies of literature, some recent studies have proposed some methods of risk evaluation for not only e-business projects. Koohathongsumrit and Meethom32, developed an integrated approach involving fuzzy risk assessment model and DEA for route selection in multimodal transportation networks. Ildarabadi and Alamatian33, proposed a new function for evaluating the financial risk of construction projects. Their proposed function is based on the Monte Carlo method. They applied the proposed risk evaluation approach to the construction industry of Iran. Song et al.34 performed a study on risk evaluation of information technology outsourcing projects. They proposed an integrated approach which considers both risk strength and interactions among risk factors. Chen35, performed a study on the risk evaluation of urban wastewater treatment projects. They proposed a study based on an improved fuzzy cognitive map. As an advantage of the proposed method, the information conflict of fuzzy evaluation among experts was solved. The study of Zhang et al.36 also is about risk evaluation of projects. They proposed an approach based on improved failure mode and effect analysis method and group decision-making technique for risk evaluation of utility tunnels construction projects. In addition to the above-mentioned approaches the data envelopment analysis models also have many applications in real-world problems but still have not been applied for risk evaluation purposes particularly for case of e-business projects seriously. For example, their applications or evaluating rural comprehensive health service centers37, performance evaluation and ranking of personnels in organizations38,39, performance evaluation of hospitals13, selection problems of energy sector40,41, etc. can be of the instances. Motivation and novelties Based on the previous explanations in this section, evaluation and prioritizing the potential risks of an e-business project is an important problem and has gained attention in literature. This study presents a novel extension of data envelopment analysis to an uncertain environment in order to evaluate and prioritize the potential risks of e-business projects. In summary, in this study the risk evaluation and ranking problem of e-business projects are considered. Several pre-determined potential risks of e-business projects are determined and their effects on project success factors are assessed by experts in the field. The evaluations are used to rank the risks by proposing a novel extension of the Tchebycheff norm model of DEA. The below points describe the research gaps, motivations, novelties, and findings of this study. As an important problem with high degrees of uncertainty, the risk evaluation and ranking problem of e-business projects has not been tackled in an uncertain environment with belief-degree-based uncertainty in literature. As an important problem with high degrees of uncertainty, the risk evaluation and ranking problem of e-business projects has not been tackled in an uncertain environment with belief-degree-based uncertainty in literature. There is no serious study in literature to apply DEA models to the risk evaluation and ranking problem of e-business projects. There is no serious study in literature to apply DEA models to the risk evaluation and ranking problem of e-business projects. In this study for the first time the risk evaluation and ranking problem of e-business projects is modeled as a data envelopment analysis model. In this study for the first time the risk evaluation and ranking problem of e-business projects is modeled as a data envelopment analysis model. Compared to literature, a wide range of risks from various aspects is considered to be evaluated and ranked. Compared to literature, a wide range of risks from various aspects is considered to be evaluated and ranked. Various aspects of projects’ success criteria are considered to evaluate the impact of each risk. In order to evaluate the impact of the risks on the success criteria, the experts of the field with experience in e-business project teams are invited to evaluate the impacts linguistically. Various aspects of projects’ success criteria are considered to evaluate the impact of each risk. In order to evaluate the impact of the risks on the success criteria, the experts of the field with experience in e-business project teams are invited to evaluate the impacts linguistically. In order to respect the uncertain nature of the risks in project management, especially e-business projects, we use the equivalent belief-degree-based uncertain values of the obtained linguistic evaluations. Belief-degree-based uncertainty is useful for cases where no historical data exists this is its advantage against fuzzy sets and numbers as in fuzzy sets and numbers the historical data is required to estimate the fuzzy value of an uncertain event. Therefore, belief-degree-based uncertainty is quite suitable for the case of estimating the impact of the potential risks on the e-business project success criteria. In order to respect the uncertain nature of the risks in project management, especially e-business projects, we use the equivalent belief-degree-based uncertain values of the obtained linguistic evaluations. Belief-degree-based uncertainty is useful for cases where no historical data exists this is its advantage against fuzzy sets and numbers as in fuzzy sets and numbers the historical data is required to estimate the fuzzy value of an uncertain event. Therefore, belief-degree-based uncertainty is quite suitable for the case of estimating the impact of the potential risks on the e-business project success criteria. In order to evaluate and rank the potential risks, the Tchebycheff norm model of DEA (see Balf et al.42 is applied. This model is quite suitable for evaluation and ranking of decision-making units (DMUs). As another novelty, in this study for the first time, this model is extended to a belief-degree-based uncertain environment. In order to evaluate and rank the potential risks, the Tchebycheff norm model of DEA (see Balf et al.42 is applied. This model is quite suitable for evaluation and ranking of decision-making units (DMUs). As another novelty, in this study for the first time, this model is extended to a belief-degree-based uncertain environment. To approaches such as chance-constrained model and expected-value model are developed to obtain equivalent crisp forms for the proposed uncertain Tchebycheff norm model. To approaches such as chance-constrained model and expected-value model are developed to obtain equivalent crisp forms for the proposed uncertain Tchebycheff norm model. The crisp models of the proposed uncertain Tchebycheff norm model are used to evaluate and rank the potential risks of e-business projects. Based on the obtained results, the most important risk is “Introducing new technology” which could be considered by the managers. In addition, an extensive sensitivity analysis and comparison study is done, and the obtained results are discussed from managerial points of view. The crisp models of the proposed uncertain Tchebycheff norm model are used to evaluate and rank the potential risks of e-business projects. Based on the obtained results, the most important risk is “Introducing new technology” which could be considered by the managers. In addition, an extensive sensitivity analysis and comparison study is done, and the obtained results are discussed from managerial points of view. Structure of the paper This paper is structured as in some sections. Section 2 represents some basic concepts about belief-degree-based uncertainty and data envelopment analysis. Section 3 describes the evaluation and ranking problem of the potential risks of e-business projects as a case study. Section 4 develops an uncertain scheme for the Tchebycheff norm model of data envelopment analysis. Section 5 represents the results obtained for the case study. Section 6 is the conclusion of the paper. Basic concepts In this section, some basic concepts related to the belief-degree-based uncertainty theory and the basic models of data envelopment analysis are presented. Uncertainty theory This sub-section presents the definitions, preliminary concepts, and symbols of uncertainty theory that are necessary for the subsequent analysis. All the concepts are taken from Liu43. Definition 1 Let be a non-empty set and L be a -algebra on . The function is defined as the belief degree function of the uncertain event (where ). The function must satisfy the following conditions: •. where is the complement for . where is the complement for . For a countable sequence where the inequality holds. For a countable sequence where the inequality holds. For infinite uncertainty spaces where the k-th one is shown by , the uncertain function is defined, where is an uncertain event from . For infinite uncertainty spaces where the k-th one is shown by , the uncertain function is defined, where is an uncertain event from . Definition 2 The uncertain variable is a function of the uncertainty space . For any Borel set B of the real numbers, the set is an uncertain event. Definition 3 The uncertainty distribution is defined as where. Definition 4 The expected value of is defined as , provided that at least one of the two integrals is finite. Definition 5 The uncertainty distribution is a regular uncertainty distribution that is continuous and strictly increasing with respect to x. It satisfies the following conditions: • • • Given that the condition is satisfied for the regular uncertainty distribution , it follows that its inverse function is defined on the interval . Definition 6 The uncertain variable in the uncertainty space , is positive if, . Definition 7 The uncertain variables from the Borel sets are said to be independent if, Theorem 1 Let be independent uncertain variables with regular uncertainty distributions , respectively. If the function is strictly increasing on and strictly decreasing on , then the uncertain variable has the inverse uncertainty distribution, Definition 8 The expected value of the uncertain variable with a regular uncertainty distribution is calculated as. Theorem 2 Let be independent uncertain variables with regular uncertainty distributions , respectively. If the function is strictly increasing on and strictly decreasing on , then the uncertain variable has the expected value of, Theorem 3 Given and as independent uncertain variables, then for any real numbers a and b, . Definition 9 A linear type uncertain variable has the following linear uncertainty distribution denoted by , where are real numbers. The inverse uncertain distribution of the linear uncertain variable is . The expected value of the linear uncertain variable is calculated by . Definition 10 An uncertain variable of zigzag type, denoted by , has the following zigzag uncertainty distribution, denoted by , where are real numbers. It is evident that the zigzag uncertainty distribution is of a regular type, and that the inverse uncertainty distribution of is represented by the following expression: The expected value of the zigzag uncertain variable is calculated by . Theorem 4 Let be an uncertain variable with a continuous uncertainty distribution . Then, for any real number x, we have, and . Lemma 1 If a and t are positive real numbers, is an independent uncertain variable with uncertain distribution , and r is a chance level. Then, is true if and only if . Ranking DMUs by the Tchebycheff norm model In this section, the classical model for ranking DMUs considering the Tchebycheff norm is presented. This model has been introduced by Balf et al.42. The model is presented as below. The most important advantage of using the Tchebycheff norm is that the obtained model is always feasible where the available ranking-based DEA models such as Andersen and Petersen44 are sometimes infeasible (see also Tian et al.45, Wang et al.46, and Iqbal et al.47. On the other hand, the proposed model is independent of the orientations such as input-oriented and output-oriented forms (see also Li and Gao48 and Huang and Chen49. subject to In the above-mentioned model, the below notations are defined. is the number of DMUs. is the number of DMUs. is the number of input-based criteria. is the number of input-based criteria. is the number of output-based criteria. is the number of output-based criteria. is the index used for DMUs. is the index used for DMUs. is the index used for the DMU which its efficiency value is being obtained. is the index used for the DMU which its efficiency value is being obtained. is the index of the input-based criteria. is the index of the input-based criteria. is the index of the output-based criteria. is the index of the output-based criteria. is the performance of DMU in input-based criterion . is the performance of DMU in input-based criterion . is the performance of DMU in output-based criterion . is the performance of DMU in output-based criterion . is a variable coefficient to determine the projected point of DMU while evaluating DMU . is a variable coefficient to determine the projected point of DMU while evaluating DMU . is a variable representing the score obtained for DMU . is a variable representing the score obtained for DMU . This model is used to measure the score of each DMU based on its performances in the given set of input-based and output-based criteria. The model is solved for each DMU and the score of each DMU is obtained. Then, the DMUs are ranked based on the descending order of their score values. Ranking problem of the potential risks of e-business projects (case study) As mentioned earlier, the risks of an e-business project play a very crucial and important role in the success of that project. Generally, in most of the e-business projects the effect of risks is very important and significantly affects the project. In this section a case study is considered with the aim of studying the effect of risks on e-business projects. For this aim, a three-stage procedure is followed to discover the success criteria of e-business projects, the set of risks affecting e-business projects, and the impact (performance) of each risk at each success criterion of e-business projects. These stages are summarized by Fig. 1. The stages to obtain the required data for studying the effect of risks of e-business projects. Stage potential risks determination Based on the literature of the field, the potential risks of e-business projects are determined. These risks are detailed by Table 1. Table 1Details of the potential risks of the e-business projects.Risk codeRisk classificationRisk descriptionRelated referencesRS1TechnologicalIntroducing new technologyStoehr7, Stevens and Fowell6RS2Site or network overload and disruptionZirakja and Samizadeh28RS3Unauthorized access by hackersZirakja and Samizadeh28RS4MarketChange in market demand rateStoehr7RS5FinancialInsufficient cash flowStoehr7, Zirakja and Samizadeh28RS6Risk of financial failureStoehr7RS7EnvironmentalChanging the goals of the projectZirakja and Samizadeh28RS8Cultural riskKhokhar et al.23RS9TechnicalReengineering riskStoehr7, Khokhar et al.23RS10Outsourcing riskStoehr7, Khokhar et al.23RS11Misunderstanding of requirements by developersStevens and Fowell6RS12Inadequate backup systemsZirakja and Samizadeh28RS13Project teamInsufficient user participation in the projectStevens and Fowell6RS14Lack of required knowledge/skills in project personnelStevens and Fowell6RS15Lack of teamwork among the groups involved in the projectStoehr7 Details of the potential risks of the e-business projects. Stage evaluation criteria determination According to the literature of project management, there are several success criteria for a project that can be influenced by the potential risks of Stage 1. These criteria are project time, project cost, project quality, project team learning, and project scope (see Stoehr7; Stevens and Fowell6; Zirakja and Samizadeh28. In addition, the degree of impact of each risk can be associated with the occurrence probability of that risk. Therefore, the probability of each risk is considered as another criterion that can be used to evaluate the risks. In this study, as a DEA based solution approach is used to evaluate the risks, the project success criteria are considered as output-based criteria and probability of each risk is considered as input-based criterion. Therefore, the considered criteria are summarized by Table 2. Here, occurrence probability of the risks is considered as an input criterion because it can directly affect output performance. The greater risk probabilities consume managerial resources and cause disruptions. It may also increase project time and cost and reduce quality. In general organizations reduce this input criterion to protect their outputs. Therefore, it could be essential for efficiency evaluation by DEA models. Table 2The evaluation criteria considered to analyze the potential risks.Criteria codeCriteria titleCriteria typeCR1Occurrence probabilityInputCR2Project timeOutputCR3Project costOutputCR4Project qualityOutputCR5Project team learningOutputCR6Project scopeOutput The evaluation criteria considered to analyze the potential risks. Stage obtaining the impact of each risk on each criterion In this stage the performance of each risk in each criterion is obtained as a decision matrix. In order to obtain this matrix, the opinions of some experts of the field are used. In order to obtain such decision matrix, the below steps are followed. Step 1. Five experts in the field are selected to evaluate the performance of each risk in each criterion. All of these experts have experience of working as a member of project team in more than two online shop platform development projects. The details of these experts are presented by Table 3. This is notable to mention that there is a degree of consistency among these experts as all of them have at least four years of experience in online shop platform development projects. Step 1. Five experts in the field are selected to evaluate the performance of each risk in each criterion. All of these experts have experience of working as a member of project team in more than two online shop platform development projects. The details of these experts are presented by Table 3. This is notable to mention that there is a degree of consistency among these experts as all of them have at least four years of experience in online shop platform development projects. Step 2. Each expert is asked to evaluate the impact (performance) of each risk on each criterion by the linguistic terms “extremely low (EL)”, “very low (VL)”, ‘low (L)”, “average (A)”, “high (H)”, “very high (VH)”, and “extremely high (EH)”. Step 2. Each expert is asked to evaluate the impact (performance) of each risk on each criterion by the linguistic terms “extremely low (EL)”, “very low (VL)”, ‘low (L)”, “average (A)”, “high (H)”, “very high (VH)”, and “extremely high (EH)”. Step 3. The linguistic evaluations obtained by Step 2, are converted to equivalent zigzag type uncertain variables by the rules of Table 4. Step 3. The linguistic evaluations obtained by Step 2, are converted to equivalent zigzag type uncertain variables by the rules of Table 4. Step 4. If the zigzag uncertain variables of risk in criterion are shown by , , …, , the below operator is used to calculate their integrated zigzag uncertain variable (). In this equation, the notation is the importance weight of Expert 1, etc. Step 4. If the zigzag uncertain variables of risk in criterion are shown by , , …, , the below operator is used to calculate their integrated zigzag uncertain variable (). In this equation, the notation is the importance weight of Expert 1, etc. Table 3The details of the experts who evaluate the performance of the risks.ExpertAcademic degreeMajorExperience (years)FieldExpert 1M.Sc.Management4Online shop platform developmentExpert 2M.Sc.Software Engineering5Online shop platform developmentExpert 3Ph.D.Software Engineering5Online shop platform developmentExpert 4Ph.D.Information Technology7Online shop platform developmentExpert 5B.Sc.Software Engineering7Online shop platform development The details of the experts who evaluate the performance of the risks. Table 4The equivalent zigzag type uncertain variables of the linguistic terms.Linguistic termEquivalent zigzag type uncertain variableExtremely low (EL) Very low (VL) Low (L) Average (A) High (H) Very high (VH) Extremely high (EH) The equivalent zigzag type uncertain variables of the linguistic terms. In order to apply the above-mentioned steps and obtain the final decision matrix of the risks and criteria, first, the followings are considered. There are 15 risks, therefore, and . There are 15 risks, therefore, and . There is 1 input-based criterion, therefore, and . There is 1 input-based criterion, therefore, and . There are 5 output-based criteria, therefore, and . There are 5 output-based criteria, therefore, and . Then, according to Step 2, the evaluations of the experts are obtained. These evaluations can be. seen in Table 5. Table 5The linguistic evaluations of the experts for the case study.CriteriaExpertRisksRS1RS2RS3RS4RS5RS6RS7RS8RS9RS10RS11RS12RS13RS14RS15CR1 (input)Expert 1ELLHAEHEHAEHVHEHEHAVLVLLExpert 2ELLAALVHLHHAHHVHVLHExpert 3VLLLHHVHVHALAAELVLVLVHExpert 4LHVLEHLLHVHVHEHHELVHVHEHExpert 5ELHLEHVHHAVHVHEHEHHAVHHCR2 (output)Expert 1VHVHLVLAAVHVLVHHLVLVHAHExpert 2HEHEHVHVHVHEHHVHEHHVHVHHHExpert 3EHVHEHAEHEHEHVHVHVHVHVHVHVHVHExpert 4HELHELEHEHEHLAAVHVHVHEHVHExpert 5HVHALVHHVHLALLAVHVHHCR3 (output)Expert 1VHHVHVHHAVHVLVHVHVLLAVHVHExpert 2VHEHVHVHHVHEHLHVHVHHHVHHExpert 3EHEHEHAEHEHEHEHVHHEHEHEHEHEHExpert 4LELLAEHEHEHLAVHHHVHEHAExpert 5HVHEHHEHEHHLVHAAAAEHACR4 (output)Expert 1HVHVHVHEHEHHELHAHVLVHEHVHExpert 2HEHVHVHEHVHAHAAVHHHVHVHExpert 3EHEHEHHVHHEHAEHLHVHVHAVHExpert 4HELAAVHVHVHLHVHEHVLAEHHExpert 5VHHHLHVHLVLLLAVLVLHACR5 (output)Expert 1AAHLLLLVLVLLVLLVLALExpert 2VHVHHVHEHVHLVHHVHHAHHHExpert 3EHELHVHHVHEHVLVHVLVLVLVHVHVHExpert 4AELLLLHALLHVHVLVLHEHExpert 5VHAAVLAVLLHLLLLAHHCR6 (output)Expert 1LVLLLVHEHEHELVLAVLLAVHLExpert 2VHVHEHVHEHHVLVHEHAHLHHVHExpert 3EHAVHHEHVHEHLVHLLLHHHExpert 4LELLVLEHEHEHLVHEHVHVHVHEHEHExpert 5AHEHLHLHLLVLHVHAAH The linguistic evaluations of the experts for the case study. In continue, applying Step 3 and Step 4, the integrated zigzag uncertain decision matrix of the risks and criteria is obtained. This is notable that in the procedure of Step 4, the experts are weighted equally. The integrated decision matrix with zigzag uncertain variables obtained for the linguistic evaluations of Table 5 is depicted by Table 6. Table 6The integrated zigzag uncertain decision matrix of the risks and criteria.RisksCriteriaCR1 (input)CR2 (output)CR3 (output)CR4 (output)CR5 (output)CR6 (output)RS1 RS2 RS3 RS4 RS5 RS6 RS7 RS8 RS9 RS10 RS11 RS12 RS13 RS14 RS15 The integrated zigzag uncertain decision matrix of the risks and criteria. The belief-degree-based uncertain Tchebycheff norm model As the data of the case study of Sect. 3 is of zigzag type uncertain variables, in this section the classical Tchebycheff norm model is extended to an uncertain environment in order to evaluate and rank the risks (DMUs) of the e-business projects. For this aim, the below notations are considered in advance. : The independent uncertain variable of DMU in input . : The independent uncertain variable of DMU in input . : The independent uncertain variable of DMU in output . : The independent uncertain variable of DMU in output . : The regular uncertain distribution of . : The regular uncertain distribution of . : The regular uncertain distribution of . : The regular uncertain distribution of . According to these notations, the uncertain form of the Tchebycheff norm model (6)-(9) is shown as the below model. subject to As the uncertain model (11)-(14) contains uncertain parameters in its constraints set, this model becomes solvable by following two phases such as (1) converting the uncertain model to a crisp form and (2) solving the obtained crisp form by any optimization solver. In order to obtain a crisp form for the uncertain model (11)-(14), two approaches of the literature can be used. According to Liu43, these approaches for obtaining a crisp form for the uncertain Tchebycheff norm model (11)-(14) are as below, Expected-value model (EVM), Expected-value model (EVM), Chance-constrained model (CCM). Chance-constrained model (CCM). These approaches are used in the rest of this section to obtain two equivalent crisp forms for the uncertain Tchebycheff norm model (11)-(14). Equivalent expected-value model (EVM) According to Liu43, the expected-value measure of an uncertain event is used to obtain its equivalent crisp form. By applying this measure, the below form of the uncertain Tchebycheff norm model (11)-(14) is obtained. subject to In continue, Theorem 7 and Corollary 1 are used to convert the expected value formulation (15)-(18) to an equivalent crisp form. Theorem 7 The uncertain model (15)-(18) is equivalent to the below formulation. subject to Proof Constraint (16) is converted to constraint (20) for by the below procedure using the definitions and theorems of Sect. 2. The conversion of constraint (17) to constraint (21) is done in similar way. Corollary 1 As the data of the case study are independent uncertainty variables with zigzag type regular uncertain distribution (e.g. , etc.), the below crisp model is obtained for the uncertain model (19)-(22). subject to Equivalent chance-constrained model (CCM) The chance-constrained approach is another effective method for converting an uncertain event to a crisp form43. In this approach the chance of occurring an uncertain event should be greater than or equal to a confidence level value between 0 and 1. Therefore, the below chance-constrained model is introduced instead of the uncertain Tchebycheff norm model (11)-(14). subject to In this model, the confidence level takes a value from the interval according to the preferences of the decision makers. Now, the uncertain model (28)-(31) is converted to a crisp form using Theorem 8, Corollary 2, and Corollary 3. Theorem 8 The uncertain model (28)-(31) is equivalent to the below formulation. subject to Proof The constraint (constraint (29)) is converted to the constraint. (constraint (33)) for using Theorem 4 and Lemma 1 defined in Sect. 2. The conversion of constraint (30) to constraint (34) is done in similar way. . Corollary 2 As the data of the case study are independent uncertainty variables with zigzag type regular uncertain distribution (e.g. , etc.), the below crisp model is obtained for the uncertain model (32)-(35) where . subject to Corollary 3 As the data of the case study are independent uncertainty variables with zigzag type regular uncertain distribution (e.g. , etc.), the below crisp model is obtained for the uncertain model (32)-(35) where . subject to General framework for applying the equivalent crisp models to the case study In order to apply the proposed equivalent crisp models of the uncertain Tchebycheff norm model (11)-(14) to the case study of Sect. 2, the general framework of Fig. 2 is followed. General framework for applying the proposed equivalent. crisp models to the case study. Computational experiments In this section the crisp models obtained for the uncertain Tchebycheff norm model (11)-(14) (the models presented by Corollary 1, Corollary 2, and Corollary 3) are used to rank the potential risks of the e-business projects according to the data of the case study of Sect. 3. All the models are coded in GAMS and solved by its CPLEX solver. The results obtained, sensitivity analysis and comparison, and some discussions are presented by the rest of this section. Results for the case study The obtained crisp forms of the uncertain Tchebycheff norm model (11)-(14) (the models (24)-(27), (36)-(39), and (40)-(43)) are solved for the data of Table 6 in order to evaluate and rank the potential risks of the e-business projects of Sect. 2. The crisp formulation EVM (24)-(27) has no confidence level, therefore, it needs only one experiment to obtain the ranking of the risks. But the formulations of the CCM (36)-(39) and CCM (40)-(43) crisp models contain some confidence level coefficients in their constraints. The range of confidence level values of the CCM (36)-(39) is the interval of and the range of confidence level values of the CCM (40)-(43) is the interval of . Therefore, the experiments of Table 7 are defined. Table 7Details of the experiments on the case study.ExperimentsConfidence levelCrisp modelAssociated rankingExp. 1-EVM (24)-(27)Ranking 1Exp. 20CCM (36)-(39)Ranking 2Exp. 30.1CCM (36)-(39)Ranking 3Exp. 40.2CCM (36)-(39)Ranking 4Exp. 50.3CCM (36)-(39)Ranking 5Exp. 60.4CCM (36)-(39)Ranking 6Exp. 70.5CCM (40)-(43)Ranking 7Exp. 80.6CCM (40)-(43)Ranking 8Exp. 90.7CCM (40)-(43)Ranking 9Exp. 100.8CCM (40)-(43)Ranking 10Exp. 110.9CCM (40)-(43)Ranking 11Exp. 121CCM (40)-(43)Ranking 12 Details of the experiments on the case study. According to the framework of Fig. 2, each experiment is executed by the data of Table 6 and as an output, the score of each risk is obtained. The score values of the risks according to the experiments of Table 7 are reported by Table 8. Table 8The scores obtained for the risks in each experiment.RisksObtained scoreExp. 1Exp. 2Exp. 3Exp. 4Exp. 5Exp. 6Exp. 7Exp. 8Exp. 9Exp. 10Exp. 11Exp. 12RS11.45740.90971.03681.17971.35561.57141.45741.69231.96302.27502.63413.0471RS20.48380.27590.32140.37040.42310.48000.48380.57220.67350.78960.92251.0746RS30.59180.29630.34620.40000.45830.59900.59180.70330.83190.98021.15091.3472RS40.35810.21620.25000.28570.32350.36360.35810.42100.49230.57330.66510.7692RS50.54850.23530.34850.54690.87100.05880.54850.64410.75290.87671.01751.1776RS60.44640.21620.25000.28570.37580.53910.44640.52480.61400.71550.83090.9620RS70.52250.25000.29030.33330.62760.02780.52250.61660.72410.84710.98741.1477RS80.27600.21050.24320.27780.31430.35290.27600.32980.39160.46270.54420.6376RS90.40410.22220.25710.29410.33330.37500.40410.47730.56090.65640.76530.8895RS100.36080.20510.23680.27030.30560.34290.36080.42270.49280.57200.66160.7629RS110.36050.21050.24320.27780.31430.35290.36050.42230.49230.57140.66090.7620RS120.51760.30770.36000.41670.47830.54550.51760.61970.73810.87551.03451.2183RS130.51120.26670.31030.35710.40740.46150.51120.60140.70430.82160.95531.1127RS140.61670.28570.33330.38460.66400.01350.61670.72610.85110.99391.15721.3438RS150.43190.22220.25710.29410.33330.43060.43190.50910.59710.69740.81170.9419 The scores obtained for the risks in each experiment. In addition, for each experiment a ranking of the risks is obtained by sorting the risks according to the descending order of their score values obtained by that experiment. Therefore, 12 rankings are obtained and reported by Table 9 where Ranking 1 is associated with the score values obtained by Exp. 1, etc. Table 9The rankings obtained for the risks in the performed experiments.RisksRanking 1Ranking 2Ranking 3Ranking 4Ranking 5Ranking 6Ranking 7Ranking 8Ranking 9Ranking 10Ranking 11Ranking 12RS1111111111111RS2856675888888RS3334462333332RS414111111129141413121212RS54832213444455RS6912121294999999RS75788414566666RS8151313131310151515151515RS911999108111111111111RS10121515151512121212131313RS11131414141411131314141414RS12622353655544RS13767786777777RS142455315222223RS1510101010117101010101010 The rankings obtained for the risks in the performed experiments. According to the obtained rankings od Table 9, the first risk “RS1” obtains the first ranking in all experiments. Therefore, the effect of this risk on e-business projects is more than other risks. For other risks, different rankings are obtained among the experiments performed. Some sensitivity analysis and comparisons are done for the obtained rankings by the next sub-sections. Sensitivity analysis The score values of Table 8 for Exp. 2 till Exp.11 are obtained by the CCM crisp formulations. Therefore, these values are plotted by Fig. 3 to show the effect of confidence level value. It can be seen that except a few cases, generally by increasing the confidence level value, the score values in all risks are increased. The formulation used for confidence levels 0 till 0.4 is different than the formulation used for confidence levels 0.5 till 1. Therefore, a decrease happens in the score values between the confidence level values 0.4 and 0.5. Graph of the score values obtained for the risks by CCM crisp formulation. over different confidence level values. The graph of rankings obtained for the risks by all experiments is depicted by Fig. 4. This graph can be used to show the sensitivity of the rankings over the confidence level value and the crisp conversion approaches such as the EVM and CCM approaches. It can be seen from the graph that except Ranking 1 and 7 (which are similar and obtained by EVM approach and CCM approach and confidence level value of 0.5 respectively) and Ranking 6 (obtained by CCM approach and confidence level value of 0.4) other rankings have more or less a degree of similarity. To be more detailed, the obtained rankings for the risks can be compared to discover their degree of similarity. As there are 12 rankings for the risks reported by Table 9, correlation value of any pair of the rankings can be calculated as a measure of similarity of the rankings. The correlation values of all pairs of the rankings are calculated and reported by Table 10. Graph of the rankings obtained for the risks by all experiments (rankings of Table 9). Table 10The correlation coefficient values between all pairs of the obtained rankings.Ranking123456789101112Ranking110.8460.8640.8710.9320.20010.9960.9930.9860.9790.975210.9460.9250.8500.4610.8460.8640.8750.8890.9110.914310.9960.9140.3790.8640.8860.8960.9110.9140.918410.9250.3430.8710.8890.9000.9140.9110.914510.1540.9320.9290.9360.9460.9360.925610.2000.2390.2460.2570.2930.339710.9960.9930.9860.9790.975810.9960.9890.9860.982910.9960.9930.9891010.9960.9931110.996121 The correlation coefficient values between all pairs of the obtained rankings. Ranking 1, which is obtained by the EVM approach, has high degree of similarity with other rankings as its correlation values with other rankings are greater than 0.8 except with Ranking 6 which is 0.2. On the other hand, Ranking 6, which is obtained by the CCM approach (confidence level value of 0.4) has a few similarities with each of the obtained rankings as the correlation values related to this ranking are less than 0.461. In summary, the obtained rankings have high degree or robustness as most of the rankings have high correlation values. Further than the above-mentioned sensitivity analysis performed on the proposed approaches, we study another sensitivity analysis on the CCM approach with confidence level of 0.5 for the cases that the importance weight of the experts are different. For this aim, we define five experiments and in each experiment the importance weight for one of the experts is considered as 0.6 and for others the weight value of 0.1 is considered. Therefore, the below experiments are defined. Experiment 1: and . Experiment 1: and . Experiment 1: and . Experiment 1: and . Experiment 1: and . Experiment 1: and . Experiment 1: and . Experiment 1: and . Experiment 1: and . Experiment 1: and . For each experiment applying its importance weight values and Eq. (10) the decision matrix of case study is obtained. Then the proposed CCM approach with confidence level of 0.5 is applied to evaluate and rank the risks of Sect. 3. Therefore, the rankings of Table 11 are obtained. In addition, to compare the obtained rankings, their correlation values are calculated and reported by Table 12. Table 11The rankings obtained for the e-business projects risks by the proposed CCM approach for different experts’ importance weight values.RisksExperiment 1Experiment 2Experiment 3Experiment 4Experiment 5RS111111RS248888RS335335RS41415151414RS589449RS663563RS751114511RS81514111515RS994994RS101312121312RS111213131113RS121177126RS1376677RS1422222RS151010101010 The rankings obtained for the e-business projects risks by the proposed CCM approach for different experts’ importance weight values. Table 12The correlation coefficient values between all pairs of the obtained rankings.RankingExperiment 1Experiment 2Experiment 3Experiment 4Experiment 5RankingExperiment 110.8000.7320.9390.789Experiment 210.8640.7640.992Experiment 310.7670.846Experiment 410.750Experiment 51 The correlation coefficient values between all pairs of the obtained rankings. According to the correlation values of Table 12, experiments 1 and 2, and experiments 2 and 5 have highest similarity among all pairs. This means that when the weight of 0.6 is considered for experts 1 and 4 or experts 2 and 5, the obtained rankings are approximately similar. On the other hand the least similarity with correlation value of 0.732 is obtained for the rankings of experiments 1 and 3. Comparative study In this section a comparative study is performed to compare the proposed approach with some deterministic approaches of literature. As there is no serious study in literature to apply ranking based DEA approaches for uncertain environment, we apply some deterministic approach. These approaches are as below. Deterministic Tchebycheff norm DEA model (6)-(9). Deterministic Tchebycheff norm DEA model (6)-(9). The approach of Anderson and Peterson (1993). The approach of Anderson and Peterson (1993). The L1-norm model50. The L1-norm model50. To execute each of the above deterministic models, the uncertain decision matrix of Table 6 is converted to a crisp decision matrix using the expected value formulation of Definition 10. The results for the above-mentioned deterministic approaches are obtained, and the obtained rankings are reported by Table 13. To compare these rankings with the rankings obtained by the approaches of this study the correlation values of the rankings of Tables 9 and 13 are calculated and reported by Table 14. Table 13The rankings obtained for the e-business projects risks by classical approaches.RisksDeterministic model (6)-(9)Anderson and Peterson (1993)L1-norm model50RS11111RS2868RS3314RS414814RS54112RS6929RS76116RS815915RS911711RS10121012RS1113913RS12545RS13737RS142113RS1510510 The rankings obtained for the e-business projects risks by classical approaches. Table 14The correlation coefficient values between all pairs of the rankings of the proposed approaches (Table 9) and the rankings of Table 13.Rankings by the proposed approachesRankings by classical approachesDeterministic model (6)-(9)Anderson and Peterson (1993)L1-norm model50Ranking 10.996−0.0320.986Ranking 20.8640.1500.832Ranking 30.8860.0460.896Ranking 40.8890.0140.907Ranking 50.929−0.1370.946Ranking 60.2390.6870.207Ranking 70.996−0.0320.986Ranking 81.0000.0000.989Ranking 90.9960.0050.986Ranking 100.9890.0140.979Ranking 110.9860.0460.968Ranking 120.9820.0910.964 The correlation coefficient values between all pairs of the rankings of the proposed approaches (Table 9) and the rankings of Table 13. According to the correlation values of Table 14, There is high similarities between the rankings obtained by the proposed approaches and the rankings of the deterministic model (6)-(9) and L1-norm model. On the other hand, the ranking obtained by the approach of Anderson and Peterson (1993) has low similarity with the rankings obtained by the proposed approaches. Discussion and managerial insights The rankings obtained by Sect. 5.2 for the risks of the e-business projects are discussed here. In order to discuss the risks, the average score values of Table 8 are calculated for each risk, then the risks are ranked based on the average score values. Therefore, the ranking of Table 15 is obtained. Table 15The average score values of all experiments and their associated ranking.RankingRiskAverage score valueRisk description1RS11.7149Introducing new technology2RS30.6914Unauthorized access by hackers3RS140.6655Lack of required knowledge/skills in project personnel4RS120.6357Inadequate backup systems5RS50.6355Insufficient cash flow6RS130.5850Insufficient user participation in the project7RS70.5747Changing the goals of the project8RS20.5725Site or network overload and disruption9RS60.5172Risk of financial failure10RS150.4965Lack of teamwork among the groups involved in the project11RS90.4699Reengineering risk12RS40.4230Change in market demand rate13RS110.4190Misunderstanding of requirements by developers14RS100.4161Outsourcing risk15RS80.3597Cultural risk The average score values of all experiments and their associated ranking. Here the most and least important risks are discussed. According to the ranking by the average score values, the most important risk is “RS1: Introducing new technology”. This risk is a popular risk in most of the e-business projects. This happens when a new competitor arrives at the market with a newer and more complete online platform. In such cases, the project owners should be ready to respond to such risk. This means that the responses such as updating the project, adding some new features, etc. can be useful to respond to the risk of introducing new technology. The second important risk of the ranking of Table 15 is “RS3: Unauthorized access by hackers”. This risk has a high impact on the success of an e-business project. Unauthorized access by hackers has many negative effects on an e-business project such as interrupting the project procedure, losing the information, etc. Therefore, the project owners and managers should be able to secure the project platform and prevent any unauthorized access by hackers. Also, the project should be ready for the event that unauthorized access by hackers happens. The least important risk according to the ranking of Table 15 is “RS8: Cultural risk”. As an e-business is performed on an online platform and the customers with different cultures can access the platform easily without facing the project owners and the business employees, the cultural risk may have a minor effect on the success of the project. Therefore, this risk may need less attention from the project owner and managers. Conclusion In this study an important problem from the e-business sector was considered. E-business projects may face several risks that can affect success of such projects. As an important issue for a project manager in this field, the potential risks should be prioritized according to their impact on the success of the project. Based on the prioritized risks, the manager can develop an effective response plan for the potential risks. For this aim a complete set of risks compared to the literature was considered and their impact on the project success criteria were determined by the experts of the field as a case study. As the potential risks of e-business projects and their impact on the success criteria of such sector have high degree of uncertainty, the evaluations obtained by the experts were converted to belief-degree-based uncertain values. Then in order to evaluate and rank the potential risks, for the first time, a DEA model was applied. For this aim, the Tchebycheff norm model for the first time was extended to a belief-degree-based uncertain environment to evaluate and rank the risks of the case study. Based on the obtained results “introducing new technology” and “cultural risk” are the most and least important risks respectively. In summary, the proposed approach has advantages such as working with belief-degree-based uncertain data which can be very useful for the situations where experts’ evaluations are important. On the other hand, the proposed framework can be extended to fuzzy theory and its various numbers to reflect the uncertainty of experts’ evaluations. Author contributions Masoomeh Shahbazi, Kazem Askarifar, and Sadegh Niroomand developed the basic idea and initial concepts of the study. Masoomeh Shahbazi and Hamidreza Taghizadeh Rajaei prepared the case study and its related data. Ali Mahmoodirad, Sadegh Niroomand, and Kazem Askarifar developed the formulations and solution methodology. Ali Mahmoodirad performed the computational experiments. Ali Mahmoodirad and Sadegh Niroomand wrote the manuscript. Data availability The data used in this study is given in the text of the paper. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data used in this study is given in the text of the paper.